index 363ce427e3fb4a6700358a8f32ae183e97aa90f9..196dfd5cc6ee3be167ce6280fb243327e77f4245 100644 (file)
-/* stb_image - v2.23 - public domain image loader - http://nothings.org/stb\r
- no warranty implied; use at your own risk\r
-\r
- Do this:\r
- #define STB_IMAGE_IMPLEMENTATION\r
- before you include this file in *one* C or C++ file to create the implementation.\r
-\r
- // i.e. it should look like this:\r
- #include ...\r
- #include ...\r
- #include ...\r
- #define STB_IMAGE_IMPLEMENTATION\r
- #include "stb_image.h"\r
-\r
- You can #define STBI_ASSERT(x) before the #include to avoid using assert.h.\r
- And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free\r
-\r
-\r
- QUICK NOTES:\r
- Primarily of interest to game developers and other people who can\r
- avoid problematic images and only need the trivial interface\r
-\r
- JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib)\r
- PNG 1/2/4/8/16-bit-per-channel\r
-\r
- TGA (not sure what subset, if a subset)\r
- BMP non-1bpp, non-RLE\r
- PSD (composited view only, no extra channels, 8/16 bit-per-channel)\r
-\r
- GIF (*comp always reports as 4-channel)\r
- HDR (radiance rgbE format)\r
- PIC (Softimage PIC)\r
- PNM (PPM and PGM binary only)\r
-\r
- Animated GIF still needs a proper API, but here's one way to do it:\r
- http://gist.github.com/urraka/685d9a6340b26b830d49\r
-\r
- - decode from memory or through FILE (define STBI_NO_STDIO to remove code)\r
- - decode from arbitrary I/O callbacks\r
- - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON)\r
-\r
- Full documentation under "DOCUMENTATION" below.\r
-\r
-\r
-LICENSE\r
-\r
- See end of file for license information.\r
-\r
-RECENT REVISION HISTORY:\r
-\r
- 2.23 (2019-08-11) fix clang static analysis warning\r
- 2.22 (2019-03-04) gif fixes, fix warnings\r
- 2.21 (2019-02-25) fix typo in comment\r
- 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs \r
- 2.19 (2018-02-11) fix warning\r
- 2.18 (2018-01-30) fix warnings\r
- 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings\r
- 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes\r
- 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC\r
- 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs\r
- 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes\r
- 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes\r
- 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64\r
- RGB-format JPEG; remove white matting in PSD;\r
- allocate large structures on the stack;\r
- correct channel count for PNG & BMP\r
- 2.10 (2016-01-22) avoid warning introduced in 2.09\r
- 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED\r
-\r
- See end of file for full revision history.\r
-\r
-\r
- ============================ Contributors =========================\r
-\r
- Image formats Extensions, features\r
- Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info)\r
- Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info)\r
- Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG)\r
- Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks)\r
- Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG)\r
- Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip)\r
- Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD)\r
- github:urraka (animated gif) Junggon Kim (PNM comments)\r
- Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA)\r
- socks-the-fox (16-bit PNG)\r
- Jeremy Sawicki (handle all ImageNet JPGs)\r
- Optimizations & bugfixes Mikhail Morozov (1-bit BMP)\r
- Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query)\r
- Arseny Kapoulkine\r
- John-Mark Allen\r
- Carmelo J Fdez-Aguera\r
-\r
- Bug & warning fixes\r
- Marc LeBlanc David Woo Guillaume George Martins Mozeiko\r
- Christpher Lloyd Jerry Jansson Joseph Thomson Phil Jordan\r
- Dave Moore Roy Eltham Hayaki Saito Nathan Reed\r
- Won Chun Luke Graham Johan Duparc Nick Verigakis\r
- the Horde3D community Thomas Ruf Ronny Chevalier github:rlyeh\r
- Janez Zemva John Bartholomew Michal Cichon github:romigrou\r
- Jonathan Blow Ken Hamada Tero Hanninen github:svdijk\r
- Laurent Gomila Cort Stratton Sergio Gonzalez github:snagar\r
- Aruelien Pocheville Thibault Reuille Cass Everitt github:Zelex\r
- Ryamond Barbiero Paul Du Bois Engin Manap github:grim210\r
- Aldo Culquicondor Philipp Wiesemann Dale Weiler github:sammyhw\r
- Oriol Ferrer Mesia Josh Tobin Matthew Gregan github:phprus\r
- Julian Raschke Gregory Mullen Baldur Karlsson github:poppolopoppo\r
- Christian Floisand Kevin Schmidt JR Smith github:darealshinji\r
- Blazej Dariusz Roszkowski github:Michaelangel007\r
-*/\r
-\r
-#ifndef STBI_INCLUDE_STB_IMAGE_H\r
-#define STBI_INCLUDE_STB_IMAGE_H\r
-\r
-// DOCUMENTATION\r
-//\r
-// Limitations:\r
-// - no 12-bit-per-channel JPEG\r
-// - no JPEGs with arithmetic coding\r
-// - GIF always returns *comp=4\r
-//\r
-// Basic usage (see HDR discussion below for HDR usage):\r
-// int x,y,n;\r
-// unsigned char *data = stbi_load(filename, &x, &y, &n, 0);\r
-// // ... process data if not NULL ...\r
-// // ... x = width, y = height, n = # 8-bit components per pixel ...\r
-// // ... replace '0' with '1'..'4' to force that many components per pixel\r
-// // ... but 'n' will always be the number that it would have been if you said 0\r
-// stbi_image_free(data)\r
-//\r
-// Standard parameters:\r
-// int *x -- outputs image width in pixels\r
-// int *y -- outputs image height in pixels\r
-// int *channels_in_file -- outputs # of image components in image file\r
-// int desired_channels -- if non-zero, # of image components requested in result\r
-//\r
-// The return value from an image loader is an 'unsigned char *' which points\r
-// to the pixel data, or NULL on an allocation failure or if the image is\r
-// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels,\r
-// with each pixel consisting of N interleaved 8-bit components; the first\r
-// pixel pointed to is top-left-most in the image. There is no padding between\r
-// image scanlines or between pixels, regardless of format. The number of\r
-// components N is 'desired_channels' if desired_channels is non-zero, or\r
-// *channels_in_file otherwise. If desired_channels is non-zero,\r
-// *channels_in_file has the number of components that _would_ have been\r
-// output otherwise. E.g. if you set desired_channels to 4, you will always\r
-// get RGBA output, but you can check *channels_in_file to see if it's trivially\r
-// opaque because e.g. there were only 3 channels in the source image.\r
-//\r
-// An output image with N components has the following components interleaved\r
-// in this order in each pixel:\r
-//\r
-// N=#comp components\r
-// 1 grey\r
-// 2 grey, alpha\r
-// 3 red, green, blue\r
-// 4 red, green, blue, alpha\r
-//\r
-// If image loading fails for any reason, the return value will be NULL,\r
-// and *x, *y, *channels_in_file will be unchanged. The function\r
-// stbi_failure_reason() can be queried for an extremely brief, end-user\r
-// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS\r
-// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly\r
-// more user-friendly ones.\r
-//\r
-// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.\r
-//\r
-// ===========================================================================\r
-//\r
-// UNICODE:\r
-//\r
-// If compiling for Windows and you wish to use Unicode filenames, compile\r
-// with\r
-// #define STBI_WINDOWS_UTF8\r
-// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert\r
-// Windows wchar_t filenames to utf8.\r
-//\r
-// ===========================================================================\r
-//\r
-// Philosophy\r
-//\r
-// stb libraries are designed with the following priorities:\r
-//\r
-// 1. easy to use\r
-// 2. easy to maintain\r
-// 3. good performance\r
-//\r
-// Sometimes I let "good performance" creep up in priority over "easy to maintain",\r
-// and for best performance I may provide less-easy-to-use APIs that give higher\r
-// performance, in addition to the easy-to-use ones. Nevertheless, it's important\r
-// to keep in mind that from the standpoint of you, a client of this library,\r
-// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all.\r
-//\r
-// Some secondary priorities arise directly from the first two, some of which\r
-// provide more explicit reasons why performance can't be emphasized.\r
-//\r
-// - Portable ("ease of use")\r
-// - Small source code footprint ("easy to maintain")\r
-// - No dependencies ("ease of use")\r
-//\r
-// ===========================================================================\r
-//\r
-// I/O callbacks\r
-//\r
-// I/O callbacks allow you to read from arbitrary sources, like packaged\r
-// files or some other source. Data read from callbacks are processed\r
-// through a small internal buffer (currently 128 bytes) to try to reduce\r
-// overhead.\r
-//\r
-// The three functions you must define are "read" (reads some bytes of data),\r
-// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end).\r
-//\r
-// ===========================================================================\r
-//\r
-// SIMD support\r
-//\r
-// The JPEG decoder will try to automatically use SIMD kernels on x86 when\r
-// supported by the compiler. For ARM Neon support, you must explicitly\r
-// request it.\r
-//\r
-// (The old do-it-yourself SIMD API is no longer supported in the current\r
-// code.)\r
-//\r
-// On x86, SSE2 will automatically be used when available based on a run-time\r
-// test; if not, the generic C versions are used as a fall-back. On ARM targets,\r
-// the typical path is to have separate builds for NEON and non-NEON devices\r
-// (at least this is true for iOS and Android). Therefore, the NEON support is\r
-// toggled by a build flag: define STBI_NEON to get NEON loops.\r
-//\r
-// If for some reason you do not want to use any of SIMD code, or if\r
-// you have issues compiling it, you can disable it entirely by\r
-// defining STBI_NO_SIMD.\r
-//\r
-// ===========================================================================\r
-//\r
-// HDR image support (disable by defining STBI_NO_HDR)\r
-//\r
-// stb_image supports loading HDR images in general, and currently the Radiance\r
-// .HDR file format specifically. You can still load any file through the existing\r
-// interface; if you attempt to load an HDR file, it will be automatically remapped\r
-// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1;\r
-// both of these constants can be reconfigured through this interface:\r
-//\r
-// stbi_hdr_to_ldr_gamma(2.2f);\r
-// stbi_hdr_to_ldr_scale(1.0f);\r
-//\r
-// (note, do not use _inverse_ constants; stbi_image will invert them\r
-// appropriately).\r
-//\r
-// Additionally, there is a new, parallel interface for loading files as\r
-// (linear) floats to preserve the full dynamic range:\r
-//\r
-// float *data = stbi_loadf(filename, &x, &y, &n, 0);\r
-//\r
-// If you load LDR images through this interface, those images will\r
-// be promoted to floating point values, run through the inverse of\r
-// constants corresponding to the above:\r
-//\r
-// stbi_ldr_to_hdr_scale(1.0f);\r
-// stbi_ldr_to_hdr_gamma(2.2f);\r
-//\r
-// Finally, given a filename (or an open file or memory block--see header\r
-// file for details) containing image data, you can query for the "most\r
-// appropriate" interface to use (that is, whether the image is HDR or\r
-// not), using:\r
-//\r
-// stbi_is_hdr(char *filename);\r
-//\r
-// ===========================================================================\r
-//\r
-// iPhone PNG support:\r
-//\r
-// By default we convert iphone-formatted PNGs back to RGB, even though\r
-// they are internally encoded differently. You can disable this conversion\r
-// by calling stbi_convert_iphone_png_to_rgb(0), in which case\r
-// you will always just get the native iphone "format" through (which\r
-// is BGR stored in RGB).\r
-//\r
-// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per\r
-// pixel to remove any premultiplied alpha *only* if the image file explicitly\r
-// says there's premultiplied data (currently only happens in iPhone images,\r
-// and only if iPhone convert-to-rgb processing is on).\r
-//\r
-// ===========================================================================\r
-//\r
-// ADDITIONAL CONFIGURATION\r
-//\r
-// - You can suppress implementation of any of the decoders to reduce\r
-// your code footprint by #defining one or more of the following\r
-// symbols before creating the implementation.\r
-//\r
-// STBI_NO_JPEG\r
-// STBI_NO_PNG\r
-// STBI_NO_BMP\r
-// STBI_NO_PSD\r
-// STBI_NO_TGA\r
-// STBI_NO_GIF\r
-// STBI_NO_HDR\r
-// STBI_NO_PIC\r
-// STBI_NO_PNM (.ppm and .pgm)\r
-//\r
-// - You can request *only* certain decoders and suppress all other ones\r
-// (this will be more forward-compatible, as addition of new decoders\r
-// doesn't require you to disable them explicitly):\r
-//\r
-// STBI_ONLY_JPEG\r
-// STBI_ONLY_PNG\r
-// STBI_ONLY_BMP\r
-// STBI_ONLY_PSD\r
-// STBI_ONLY_TGA\r
-// STBI_ONLY_GIF\r
-// STBI_ONLY_HDR\r
-// STBI_ONLY_PIC\r
-// STBI_ONLY_PNM (.ppm and .pgm)\r
-//\r
-// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still\r
-// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB\r
-//\r
-\r
-\r
-#ifndef STBI_NO_STDIO\r
-#include <stdio.h>\r
-#endif // STBI_NO_STDIO\r
-\r
-#define STBI_VERSION 1\r
-\r
-enum\r
-{\r
- STBI_default = 0, // only used for desired_channels\r
-\r
- STBI_grey = 1,\r
- STBI_grey_alpha = 2,\r
- STBI_rgb = 3,\r
- STBI_rgb_alpha = 4\r
-};\r
-\r
-#include <stdlib.h>\r
-typedef unsigned char stbi_uc;\r
-typedef unsigned short stbi_us;\r
-\r
-#ifdef __cplusplus\r
-extern "C" {\r
-#endif\r
-\r
-#ifndef STBIDEF\r
-#ifdef STB_IMAGE_STATIC\r
-#define STBIDEF static\r
-#else\r
-#define STBIDEF extern\r
-#endif\r
-#endif\r
-\r
-//////////////////////////////////////////////////////////////////////////////\r
-//\r
-// PRIMARY API - works on images of any type\r
-//\r
-\r
-//\r
-// load image by filename, open file, or memory buffer\r
-//\r
-\r
-typedef struct\r
-{\r
- int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read\r
- void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative\r
- int (*eof) (void *user); // returns nonzero if we are at end of file/data\r
-} stbi_io_callbacks;\r
-\r
-////////////////////////////////////\r
-//\r
-// 8-bits-per-channel interface\r
-//\r
-\r
-STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels);\r
-STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels);\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);\r
-STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);\r
-// for stbi_load_from_file, file pointer is left pointing immediately after image\r
-#endif\r
-\r
-#ifndef STBI_NO_GIF\r
-STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp);\r
-#endif\r
-\r
-#ifdef STBI_WINDOWS_UTF8\r
-STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input);\r
-#endif\r
-\r
-////////////////////////////////////\r
-//\r
-// 16-bits-per-channel interface\r
-//\r
-\r
-STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);\r
-STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);\r
-STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);\r
-#endif\r
-\r
-////////////////////////////////////\r
-//\r
-// float-per-channel interface\r
-//\r
-#ifndef STBI_NO_LINEAR\r
- STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);\r
- STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);\r
-\r
- #ifndef STBI_NO_STDIO\r
- STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);\r
- STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);\r
- #endif\r
-#endif\r
-\r
-#ifndef STBI_NO_HDR\r
- STBIDEF void stbi_hdr_to_ldr_gamma(float gamma);\r
- STBIDEF void stbi_hdr_to_ldr_scale(float scale);\r
-#endif // STBI_NO_HDR\r
-\r
-#ifndef STBI_NO_LINEAR\r
- STBIDEF void stbi_ldr_to_hdr_gamma(float gamma);\r
- STBIDEF void stbi_ldr_to_hdr_scale(float scale);\r
-#endif // STBI_NO_LINEAR\r
-\r
-// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR\r
-STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user);\r
-STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len);\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF int stbi_is_hdr (char const *filename);\r
-STBIDEF int stbi_is_hdr_from_file(FILE *f);\r
-#endif // STBI_NO_STDIO\r
-\r
-\r
-// get a VERY brief reason for failure\r
-// NOT THREADSAFE\r
-STBIDEF const char *stbi_failure_reason (void);\r
-\r
-// free the loaded image -- this is just free()\r
-STBIDEF void stbi_image_free (void *retval_from_stbi_load);\r
-\r
-// get image dimensions & components without fully decoding\r
-STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp);\r
-STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp);\r
-STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len);\r
-STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user);\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp);\r
-STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp);\r
-STBIDEF int stbi_is_16_bit (char const *filename);\r
-STBIDEF int stbi_is_16_bit_from_file(FILE *f);\r
-#endif\r
-\r
-\r
-\r
-// for image formats that explicitly notate that they have premultiplied alpha,\r
-// we just return the colors as stored in the file. set this flag to force\r
-// unpremultiplication. results are undefined if the unpremultiply overflow.\r
-STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply);\r
-\r
-// indicate whether we should process iphone images back to canonical format,\r
-// or just pass them through "as-is"\r
-STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert);\r
-\r
-// flip the image vertically, so the first pixel in the output array is the bottom left\r
-STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip);\r
-\r
-// ZLIB client - used by PNG, available for other purposes\r
-\r
-STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);\r
-STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);\r
-STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);\r
-STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);\r
-\r
-STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);\r
-STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);\r
-\r
-\r
-#ifdef __cplusplus\r
-}\r
-#endif\r
-\r
-//\r
-//\r
-//// end header file /////////////////////////////////////////////////////\r
-#endif // STBI_INCLUDE_STB_IMAGE_H\r
-\r
-#ifdef STB_IMAGE_IMPLEMENTATION\r
-\r
-#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \\r
- || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \\r
- || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \\r
- || defined(STBI_ONLY_ZLIB)\r
- #ifndef STBI_ONLY_JPEG\r
- #define STBI_NO_JPEG\r
- #endif\r
- #ifndef STBI_ONLY_PNG\r
- #define STBI_NO_PNG\r
- #endif\r
- #ifndef STBI_ONLY_BMP\r
- #define STBI_NO_BMP\r
- #endif\r
- #ifndef STBI_ONLY_PSD\r
- #define STBI_NO_PSD\r
- #endif\r
- #ifndef STBI_ONLY_TGA\r
- #define STBI_NO_TGA\r
- #endif\r
- #ifndef STBI_ONLY_GIF\r
- #define STBI_NO_GIF\r
- #endif\r
- #ifndef STBI_ONLY_HDR\r
- #define STBI_NO_HDR\r
- #endif\r
- #ifndef STBI_ONLY_PIC\r
- #define STBI_NO_PIC\r
- #endif\r
- #ifndef STBI_ONLY_PNM\r
- #define STBI_NO_PNM\r
- #endif\r
-#endif\r
-\r
-#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB)\r
-#define STBI_NO_ZLIB\r
-#endif\r
-\r
-\r
-#include <stdarg.h>\r
-#include <stddef.h> // ptrdiff_t on osx\r
-#include <stdlib.h>\r
-#include <string.h>\r
-#include <limits.h>\r
-\r
-#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)\r
-#include <math.h> // ldexp, pow\r
-#endif\r
-\r
-#ifndef STBI_NO_STDIO\r
-#include <stdio.h>\r
-#endif\r
-\r
-#ifndef STBI_ASSERT\r
-#include <assert.h>\r
-#define STBI_ASSERT(x) assert(x)\r
-#endif\r
-\r
-#ifdef __cplusplus\r
-#define STBI_EXTERN extern "C"\r
-#else\r
-#define STBI_EXTERN extern\r
-#endif\r
-\r
-\r
-#ifndef _MSC_VER\r
- #ifdef __cplusplus\r
- #define stbi_inline inline\r
- #else\r
- #define stbi_inline\r
- #endif\r
-#else\r
- #define stbi_inline __forceinline\r
-#endif\r
-\r
-\r
-#ifdef _MSC_VER\r
-typedef unsigned short stbi__uint16;\r
-typedef signed short stbi__int16;\r
-typedef unsigned int stbi__uint32;\r
-typedef signed int stbi__int32;\r
-#else\r
-#include <stdint.h>\r
-typedef uint16_t stbi__uint16;\r
-typedef int16_t stbi__int16;\r
-typedef uint32_t stbi__uint32;\r
-typedef int32_t stbi__int32;\r
-#endif\r
-\r
-// should produce compiler error if size is wrong\r
-typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1];\r
-\r
-#ifdef _MSC_VER\r
-#define STBI_NOTUSED(v) (void)(v)\r
-#else\r
-#define STBI_NOTUSED(v) (void)sizeof(v)\r
-#endif\r
-\r
-#ifdef _MSC_VER\r
-#define STBI_HAS_LROTL\r
-#endif\r
-\r
-#ifdef STBI_HAS_LROTL\r
- #define stbi_lrot(x,y) _lrotl(x,y)\r
-#else\r
- #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y))))\r
-#endif\r
-\r
-#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED))\r
-// ok\r
-#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED)\r
-// ok\r
-#else\r
-#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)."\r
-#endif\r
-\r
-#ifndef STBI_MALLOC\r
-#define STBI_MALLOC(sz) malloc(sz)\r
-#define STBI_REALLOC(p,newsz) realloc(p,newsz)\r
-#define STBI_FREE(p) free(p)\r
-#endif\r
-\r
-#ifndef STBI_REALLOC_SIZED\r
-#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz)\r
-#endif\r
-\r
-// x86/x64 detection\r
-#if defined(__x86_64__) || defined(_M_X64)\r
-#define STBI__X64_TARGET\r
-#elif defined(__i386) || defined(_M_IX86)\r
-#define STBI__X86_TARGET\r
-#endif\r
-\r
-#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD)\r
-// gcc doesn't support sse2 intrinsics unless you compile with -msse2,\r
-// which in turn means it gets to use SSE2 everywhere. This is unfortunate,\r
-// but previous attempts to provide the SSE2 functions with runtime\r
-// detection caused numerous issues. The way architecture extensions are\r
-// exposed in GCC/Clang is, sadly, not really suited for one-file libs.\r
-// New behavior: if compiled with -msse2, we use SSE2 without any\r
-// detection; if not, we don't use it at all.\r
-#define STBI_NO_SIMD\r
-#endif\r
-\r
-#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD)\r
-// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET\r
-//\r
-// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the\r
-// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant.\r
-// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not\r
-// simultaneously enabling "-mstackrealign".\r
-//\r
-// See https://github.com/nothings/stb/issues/81 for more information.\r
-//\r
-// So default to no SSE2 on 32-bit MinGW. If you've read this far and added\r
-// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2.\r
-#define STBI_NO_SIMD\r
-#endif\r
-\r
-#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET))\r
-#define STBI_SSE2\r
-#include <emmintrin.h>\r
-\r
-#ifdef _MSC_VER\r
-\r
-#if _MSC_VER >= 1400 // not VC6\r
-#include <intrin.h> // __cpuid\r
-static int stbi__cpuid3(void)\r
-{\r
- int info[4];\r
- __cpuid(info,1);\r
- return info[3];\r
-}\r
-#else\r
-static int stbi__cpuid3(void)\r
-{\r
- int res;\r
- __asm {\r
- mov eax,1\r
- cpuid\r
- mov res,edx\r
- }\r
- return res;\r
-}\r
-#endif\r
-\r
-#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name\r
-\r
-#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)\r
-static int stbi__sse2_available(void)\r
-{\r
- int info3 = stbi__cpuid3();\r
- return ((info3 >> 26) & 1) != 0;\r
-}\r
-#endif\r
-\r
-#else // assume GCC-style if not VC++\r
-#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))\r
-\r
-#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)\r
-static int stbi__sse2_available(void)\r
-{\r
- // If we're even attempting to compile this on GCC/Clang, that means\r
- // -msse2 is on, which means the compiler is allowed to use SSE2\r
- // instructions at will, and so are we.\r
- return 1;\r
-}\r
-#endif\r
-\r
-#endif\r
-#endif\r
-\r
-// ARM NEON\r
-#if defined(STBI_NO_SIMD) && defined(STBI_NEON)\r
-#undef STBI_NEON\r
-#endif\r
-\r
-#ifdef STBI_NEON\r
-#include <arm_neon.h>\r
-// assume GCC or Clang on ARM targets\r
-#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))\r
-#endif\r
-\r
-#ifndef STBI_SIMD_ALIGN\r
-#define STBI_SIMD_ALIGN(type, name) type name\r
-#endif\r
-\r
-///////////////////////////////////////////////\r
-//\r
-// stbi__context struct and start_xxx functions\r
-\r
-// stbi__context structure is our basic context used by all images, so it\r
-// contains all the IO context, plus some basic image information\r
-typedef struct\r
-{\r
- stbi__uint32 img_x, img_y;\r
- int img_n, img_out_n;\r
-\r
- stbi_io_callbacks io;\r
- void *io_user_data;\r
-\r
- int read_from_callbacks;\r
- int buflen;\r
- stbi_uc buffer_start[128];\r
-\r
- stbi_uc *img_buffer, *img_buffer_end;\r
- stbi_uc *img_buffer_original, *img_buffer_original_end;\r
-} stbi__context;\r
-\r
-\r
-static void stbi__refill_buffer(stbi__context *s);\r
-\r
-// initialize a memory-decode context\r
-static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len)\r
-{\r
- s->io.read = NULL;\r
- s->read_from_callbacks = 0;\r
- s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer;\r
- s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len;\r
-}\r
-\r
-// initialize a callback-based context\r
-static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)\r
-{\r
- s->io = *c;\r
- s->io_user_data = user;\r
- s->buflen = sizeof(s->buffer_start);\r
- s->read_from_callbacks = 1;\r
- s->img_buffer_original = s->buffer_start;\r
- stbi__refill_buffer(s);\r
- s->img_buffer_original_end = s->img_buffer_end;\r
-}\r
-\r
-#ifndef STBI_NO_STDIO\r
-\r
-static int stbi__stdio_read(void *user, char *data, int size)\r
-{\r
- return (int) fread(data,1,size,(FILE*) user);\r
-}\r
-\r
-static void stbi__stdio_skip(void *user, int n)\r
-{\r
- fseek((FILE*) user, n, SEEK_CUR);\r
-}\r
-\r
-static int stbi__stdio_eof(void *user)\r
-{\r
- return feof((FILE*) user);\r
-}\r
-\r
-static stbi_io_callbacks stbi__stdio_callbacks =\r
-{\r
- stbi__stdio_read,\r
- stbi__stdio_skip,\r
- stbi__stdio_eof,\r
-};\r
-\r
-static void stbi__start_file(stbi__context *s, FILE *f)\r
-{\r
- stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);\r
-}\r
-\r
-//static void stop_file(stbi__context *s) { }\r
-\r
-#endif // !STBI_NO_STDIO\r
-\r
-static void stbi__rewind(stbi__context *s)\r
-{\r
- // conceptually rewind SHOULD rewind to the beginning of the stream,\r
- // but we just rewind to the beginning of the initial buffer, because\r
- // we only use it after doing 'test', which only ever looks at at most 92 bytes\r
- s->img_buffer = s->img_buffer_original;\r
- s->img_buffer_end = s->img_buffer_original_end;\r
-}\r
-\r
-enum\r
-{\r
- STBI_ORDER_RGB,\r
- STBI_ORDER_BGR\r
-};\r
-\r
-typedef struct\r
-{\r
- int bits_per_channel;\r
- int num_channels;\r
- int channel_order;\r
-} stbi__result_info;\r
-\r
-#ifndef STBI_NO_JPEG\r
-static int stbi__jpeg_test(stbi__context *s);\r
-static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_PNG\r
-static int stbi__png_test(stbi__context *s);\r
-static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp);\r
-static int stbi__png_is16(stbi__context *s);\r
-#endif\r
-\r
-#ifndef STBI_NO_BMP\r
-static int stbi__bmp_test(stbi__context *s);\r
-static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_TGA\r
-static int stbi__tga_test(stbi__context *s);\r
-static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_PSD\r
-static int stbi__psd_test(stbi__context *s);\r
-static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc);\r
-static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp);\r
-static int stbi__psd_is16(stbi__context *s);\r
-#endif\r
-\r
-#ifndef STBI_NO_HDR\r
-static int stbi__hdr_test(stbi__context *s);\r
-static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_PIC\r
-static int stbi__pic_test(stbi__context *s);\r
-static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_GIF\r
-static int stbi__gif_test(stbi__context *s);\r
-static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp);\r
-static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_PNM\r
-static int stbi__pnm_test(stbi__context *s);\r
-static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);\r
-static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp);\r
-#endif\r
-\r
-// this is not threadsafe\r
-static const char *stbi__g_failure_reason;\r
-\r
-STBIDEF const char *stbi_failure_reason(void)\r
-{\r
- return stbi__g_failure_reason;\r
-}\r
-\r
-static int stbi__err(const char *str)\r
-{\r
- stbi__g_failure_reason = str;\r
- return 0;\r
-}\r
-\r
-static void *stbi__malloc(size_t size)\r
-{\r
- return STBI_MALLOC(size);\r
-}\r
-\r
-// stb_image uses ints pervasively, including for offset calculations.\r
-// therefore the largest decoded image size we can support with the\r
-// current code, even on 64-bit targets, is INT_MAX. this is not a\r
-// significant limitation for the intended use case.\r
-//\r
-// we do, however, need to make sure our size calculations don't\r
-// overflow. hence a few helper functions for size calculations that\r
-// multiply integers together, making sure that they're non-negative\r
-// and no overflow occurs.\r
-\r
-// return 1 if the sum is valid, 0 on overflow.\r
-// negative terms are considered invalid.\r
-static int stbi__addsizes_valid(int a, int b)\r
-{\r
- if (b < 0) return 0;\r
- // now 0 <= b <= INT_MAX, hence also\r
- // 0 <= INT_MAX - b <= INTMAX.\r
- // And "a + b <= INT_MAX" (which might overflow) is the\r
- // same as a <= INT_MAX - b (no overflow)\r
- return a <= INT_MAX - b;\r
-}\r
-\r
-// returns 1 if the product is valid, 0 on overflow.\r
-// negative factors are considered invalid.\r
-static int stbi__mul2sizes_valid(int a, int b)\r
-{\r
- if (a < 0 || b < 0) return 0;\r
- if (b == 0) return 1; // mul-by-0 is always safe\r
- // portable way to check for no overflows in a*b\r
- return a <= INT_MAX/b;\r
-}\r
-\r
-// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow\r
-static int stbi__mad2sizes_valid(int a, int b, int add)\r
-{\r
- return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add);\r
-}\r
-\r
-// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow\r
-static int stbi__mad3sizes_valid(int a, int b, int c, int add)\r
-{\r
- return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&\r
- stbi__addsizes_valid(a*b*c, add);\r
-}\r
-\r
-// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow\r
-#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)\r
-static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add)\r
-{\r
- return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&\r
- stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add);\r
-}\r
-#endif\r
-\r
-// mallocs with size overflow checking\r
-static void *stbi__malloc_mad2(int a, int b, int add)\r
-{\r
- if (!stbi__mad2sizes_valid(a, b, add)) return NULL;\r
- return stbi__malloc(a*b + add);\r
-}\r
-\r
-static void *stbi__malloc_mad3(int a, int b, int c, int add)\r
-{\r
- if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL;\r
- return stbi__malloc(a*b*c + add);\r
-}\r
-\r
-#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)\r
-static void *stbi__malloc_mad4(int a, int b, int c, int d, int add)\r
-{\r
- if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL;\r
- return stbi__malloc(a*b*c*d + add);\r
-}\r
-#endif\r
-\r
-// stbi__err - error\r
-// stbi__errpf - error returning pointer to float\r
-// stbi__errpuc - error returning pointer to unsigned char\r
-\r
-#ifdef STBI_NO_FAILURE_STRINGS\r
- #define stbi__err(x,y) 0\r
-#elif defined(STBI_FAILURE_USERMSG)\r
- #define stbi__err(x,y) stbi__err(y)\r
-#else\r
- #define stbi__err(x,y) stbi__err(x)\r
-#endif\r
-\r
-#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL))\r
-#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))\r
-\r
-STBIDEF void stbi_image_free(void *retval_from_stbi_load)\r
-{\r
- STBI_FREE(retval_from_stbi_load);\r
-}\r
-\r
-#ifndef STBI_NO_LINEAR\r
-static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp);\r
-#endif\r
-\r
-#ifndef STBI_NO_HDR\r
-static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp);\r
-#endif\r
-\r
-static int stbi__vertically_flip_on_load = 0;\r
-\r
-STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip)\r
-{\r
- stbi__vertically_flip_on_load = flag_true_if_should_flip;\r
-}\r
-\r
-static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)\r
-{\r
- memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields\r
- ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed\r
- ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order\r
- ri->num_channels = 0;\r
-\r
- #ifndef STBI_NO_JPEG\r
- if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
- #ifndef STBI_NO_PNG\r
- if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
- #ifndef STBI_NO_BMP\r
- if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
- #ifndef STBI_NO_GIF\r
- if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
- #ifndef STBI_NO_PSD\r
- if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc);\r
- #endif\r
- #ifndef STBI_NO_PIC\r
- if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
- #ifndef STBI_NO_PNM\r
- if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
-\r
- #ifndef STBI_NO_HDR\r
- if (stbi__hdr_test(s)) {\r
- float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri);\r
- return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp);\r
- }\r
- #endif\r
-\r
- #ifndef STBI_NO_TGA\r
- // test tga last because it's a crappy test!\r
- if (stbi__tga_test(s))\r
- return stbi__tga_load(s,x,y,comp,req_comp, ri);\r
- #endif\r
-\r
- return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt");\r
-}\r
-\r
-static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels)\r
-{\r
- int i;\r
- int img_len = w * h * channels;\r
- stbi_uc *reduced;\r
-\r
- reduced = (stbi_uc *) stbi__malloc(img_len);\r
- if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory");\r
-\r
- for (i = 0; i < img_len; ++i)\r
- reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling\r
-\r
- STBI_FREE(orig);\r
- return reduced;\r
-}\r
-\r
-static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels)\r
-{\r
- int i;\r
- int img_len = w * h * channels;\r
- stbi__uint16 *enlarged;\r
-\r
- enlarged = (stbi__uint16 *) stbi__malloc(img_len*2);\r
- if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");\r
-\r
- for (i = 0; i < img_len; ++i)\r
- enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff\r
-\r
- STBI_FREE(orig);\r
- return enlarged;\r
-}\r
-\r
-static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel)\r
-{\r
- int row;\r
- size_t bytes_per_row = (size_t)w * bytes_per_pixel;\r
- stbi_uc temp[2048];\r
- stbi_uc *bytes = (stbi_uc *)image;\r
-\r
- for (row = 0; row < (h>>1); row++) {\r
- stbi_uc *row0 = bytes + row*bytes_per_row;\r
- stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row;\r
- // swap row0 with row1\r
- size_t bytes_left = bytes_per_row;\r
- while (bytes_left) {\r
- size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);\r
- memcpy(temp, row0, bytes_copy);\r
- memcpy(row0, row1, bytes_copy);\r
- memcpy(row1, temp, bytes_copy);\r
- row0 += bytes_copy;\r
- row1 += bytes_copy;\r
- bytes_left -= bytes_copy;\r
- }\r
- }\r
-}\r
-\r
-#ifndef STBI_NO_GIF\r
-static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel)\r
-{\r
- int slice;\r
- int slice_size = w * h * bytes_per_pixel;\r
-\r
- stbi_uc *bytes = (stbi_uc *)image;\r
- for (slice = 0; slice < z; ++slice) {\r
- stbi__vertical_flip(bytes, w, h, bytes_per_pixel); \r
- bytes += slice_size; \r
- }\r
-}\r
-#endif\r
-\r
-static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__result_info ri;\r
- void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8);\r
-\r
- if (result == NULL)\r
- return NULL;\r
-\r
- if (ri.bits_per_channel != 8) {\r
- STBI_ASSERT(ri.bits_per_channel == 16);\r
- result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp);\r
- ri.bits_per_channel = 8;\r
- }\r
-\r
- // @TODO: move stbi__convert_format to here\r
-\r
- if (stbi__vertically_flip_on_load) {\r
- int channels = req_comp ? req_comp : *comp;\r
- stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc));\r
- }\r
-\r
- return (unsigned char *) result;\r
-}\r
-\r
-static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__result_info ri;\r
- void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16);\r
-\r
- if (result == NULL)\r
- return NULL;\r
-\r
- if (ri.bits_per_channel != 16) {\r
- STBI_ASSERT(ri.bits_per_channel == 8);\r
- result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp);\r
- ri.bits_per_channel = 16;\r
- }\r
-\r
- // @TODO: move stbi__convert_format16 to here\r
- // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision\r
-\r
- if (stbi__vertically_flip_on_load) {\r
- int channels = req_comp ? req_comp : *comp;\r
- stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16));\r
- }\r
-\r
- return (stbi__uint16 *) result;\r
-}\r
-\r
-#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR)\r
-static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)\r
-{\r
- if (stbi__vertically_flip_on_load && result != NULL) {\r
- int channels = req_comp ? req_comp : *comp;\r
- stbi__vertical_flip(result, *x, *y, channels * sizeof(float));\r
- }\r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_STDIO\r
-\r
-#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)\r
-STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide);\r
-STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default);\r
-#endif\r
-\r
-#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)\r
-STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input)\r
-{\r
- return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL);\r
-}\r
-#endif\r
-\r
-static FILE *stbi__fopen(char const *filename, char const *mode)\r
-{\r
- FILE *f;\r
-#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)\r
- wchar_t wMode[64];\r
- wchar_t wFilename[1024];\r
- if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)))\r
- return 0;\r
- \r
- if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)))\r
- return 0;\r
-\r
-#if _MSC_VER >= 1400\r
- if (0 != _wfopen_s(&f, wFilename, wMode))\r
- f = 0;\r
-#else\r
- f = _wfopen(wFilename, wMode);\r
-#endif\r
-\r
-#elif defined(_MSC_VER) && _MSC_VER >= 1400\r
- if (0 != fopen_s(&f, filename, mode))\r
- f=0;\r
-#else\r
- f = fopen(filename, mode);\r
-#endif\r
- return f;\r
-}\r
-\r
-\r
-STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp)\r
-{\r
- FILE *f = stbi__fopen(filename, "rb");\r
- unsigned char *result;\r
- if (!f) return stbi__errpuc("can't fopen", "Unable to open file");\r
- result = stbi_load_from_file(f,x,y,comp,req_comp);\r
- fclose(f);\r
- return result;\r
-}\r
-\r
-STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)\r
-{\r
- unsigned char *result;\r
- stbi__context s;\r
- stbi__start_file(&s,f);\r
- result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);\r
- if (result) {\r
- // need to 'unget' all the characters in the IO buffer\r
- fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);\r
- }\r
- return result;\r
-}\r
-\r
-STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__uint16 *result;\r
- stbi__context s;\r
- stbi__start_file(&s,f);\r
- result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp);\r
- if (result) {\r
- // need to 'unget' all the characters in the IO buffer\r
- fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);\r
- }\r
- return result;\r
-}\r
-\r
-STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp)\r
-{\r
- FILE *f = stbi__fopen(filename, "rb");\r
- stbi__uint16 *result;\r
- if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file");\r
- result = stbi_load_from_file_16(f,x,y,comp,req_comp);\r
- fclose(f);\r
- return result;\r
-}\r
-\r
-\r
-#endif //!STBI_NO_STDIO\r
-\r
-STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels)\r
-{\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);\r
-}\r
-\r
-STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels)\r
-{\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user);\r
- return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);\r
-}\r
-\r
-STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);\r
-}\r
-\r
-STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\r
- return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);\r
-}\r
-\r
-#ifndef STBI_NO_GIF\r
-STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp)\r
-{\r
- unsigned char *result;\r
- stbi__context s; \r
- stbi__start_mem(&s,buffer,len); \r
- \r
- result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp);\r
- if (stbi__vertically_flip_on_load) {\r
- stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); \r
- }\r
-\r
- return result; \r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_LINEAR\r
-static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)\r
-{\r
- unsigned char *data;\r
- #ifndef STBI_NO_HDR\r
- if (stbi__hdr_test(s)) {\r
- stbi__result_info ri;\r
- float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri);\r
- if (hdr_data)\r
- stbi__float_postprocess(hdr_data,x,y,comp,req_comp);\r
- return hdr_data;\r
- }\r
- #endif\r
- data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp);\r
- if (data)\r
- return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp);\r
- return stbi__errpf("unknown image type", "Image not of any known type, or corrupt");\r
-}\r
-\r
-STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__loadf_main(&s,x,y,comp,req_comp);\r
-}\r
-\r
-STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\r
- return stbi__loadf_main(&s,x,y,comp,req_comp);\r
-}\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp)\r
-{\r
- float *result;\r
- FILE *f = stbi__fopen(filename, "rb");\r
- if (!f) return stbi__errpf("can't fopen", "Unable to open file");\r
- result = stbi_loadf_from_file(f,x,y,comp,req_comp);\r
- fclose(f);\r
- return result;\r
-}\r
-\r
-STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)\r
-{\r
- stbi__context s;\r
- stbi__start_file(&s,f);\r
- return stbi__loadf_main(&s,x,y,comp,req_comp);\r
-}\r
-#endif // !STBI_NO_STDIO\r
-\r
-#endif // !STBI_NO_LINEAR\r
-\r
-// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is\r
-// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always\r
-// reports false!\r
-\r
-STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len)\r
-{\r
- #ifndef STBI_NO_HDR\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__hdr_test(&s);\r
- #else\r
- STBI_NOTUSED(buffer);\r
- STBI_NOTUSED(len);\r
- return 0;\r
- #endif\r
-}\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF int stbi_is_hdr (char const *filename)\r
-{\r
- FILE *f = stbi__fopen(filename, "rb");\r
- int result=0;\r
- if (f) {\r
- result = stbi_is_hdr_from_file(f);\r
- fclose(f);\r
- }\r
- return result;\r
-}\r
-\r
-STBIDEF int stbi_is_hdr_from_file(FILE *f)\r
-{\r
- #ifndef STBI_NO_HDR\r
- long pos = ftell(f);\r
- int res;\r
- stbi__context s;\r
- stbi__start_file(&s,f);\r
- res = stbi__hdr_test(&s);\r
- fseek(f, pos, SEEK_SET);\r
- return res;\r
- #else\r
- STBI_NOTUSED(f);\r
- return 0;\r
- #endif\r
-}\r
-#endif // !STBI_NO_STDIO\r
-\r
-STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user)\r
-{\r
- #ifndef STBI_NO_HDR\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\r
- return stbi__hdr_test(&s);\r
- #else\r
- STBI_NOTUSED(clbk);\r
- STBI_NOTUSED(user);\r
- return 0;\r
- #endif\r
-}\r
-\r
-#ifndef STBI_NO_LINEAR\r
-static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f;\r
-\r
-STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; }\r
-STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; }\r
-#endif\r
-\r
-static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f;\r
-\r
-STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; }\r
-STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; }\r
-\r
-\r
-//////////////////////////////////////////////////////////////////////////////\r
-//\r
-// Common code used by all image loaders\r
-//\r
-\r
-enum\r
-{\r
- STBI__SCAN_load=0,\r
- STBI__SCAN_type,\r
- STBI__SCAN_header\r
-};\r
-\r
-static void stbi__refill_buffer(stbi__context *s)\r
-{\r
- int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);\r
- if (n == 0) {\r
- // at end of file, treat same as if from memory, but need to handle case\r
- // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file\r
- s->read_from_callbacks = 0;\r
- s->img_buffer = s->buffer_start;\r
- s->img_buffer_end = s->buffer_start+1;\r
- *s->img_buffer = 0;\r
- } else {\r
- s->img_buffer = s->buffer_start;\r
- s->img_buffer_end = s->buffer_start + n;\r
- }\r
-}\r
-\r
-stbi_inline static stbi_uc stbi__get8(stbi__context *s)\r
-{\r
- if (s->img_buffer < s->img_buffer_end)\r
- return *s->img_buffer++;\r
- if (s->read_from_callbacks) {\r
- stbi__refill_buffer(s);\r
- return *s->img_buffer++;\r
- }\r
- return 0;\r
-}\r
-\r
-stbi_inline static int stbi__at_eof(stbi__context *s)\r
-{\r
- if (s->io.read) {\r
- if (!(s->io.eof)(s->io_user_data)) return 0;\r
- // if feof() is true, check if buffer = end\r
- // special case: we've only got the special 0 character at the end\r
- if (s->read_from_callbacks == 0) return 1;\r
- }\r
-\r
- return s->img_buffer >= s->img_buffer_end;\r
-}\r
-\r
-static void stbi__skip(stbi__context *s, int n)\r
-{\r
- if (n < 0) {\r
- s->img_buffer = s->img_buffer_end;\r
- return;\r
- }\r
- if (s->io.read) {\r
- int blen = (int) (s->img_buffer_end - s->img_buffer);\r
- if (blen < n) {\r
- s->img_buffer = s->img_buffer_end;\r
- (s->io.skip)(s->io_user_data, n - blen);\r
- return;\r
- }\r
- }\r
- s->img_buffer += n;\r
-}\r
-\r
-static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n)\r
-{\r
- if (s->io.read) {\r
- int blen = (int) (s->img_buffer_end - s->img_buffer);\r
- if (blen < n) {\r
- int res, count;\r
-\r
- memcpy(buffer, s->img_buffer, blen);\r
-\r
- count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);\r
- res = (count == (n-blen));\r
- s->img_buffer = s->img_buffer_end;\r
- return res;\r
- }\r
- }\r
-\r
- if (s->img_buffer+n <= s->img_buffer_end) {\r
- memcpy(buffer, s->img_buffer, n);\r
- s->img_buffer += n;\r
- return 1;\r
- } else\r
- return 0;\r
-}\r
-\r
-static int stbi__get16be(stbi__context *s)\r
-{\r
- int z = stbi__get8(s);\r
- return (z << 8) + stbi__get8(s);\r
-}\r
-\r
-static stbi__uint32 stbi__get32be(stbi__context *s)\r
-{\r
- stbi__uint32 z = stbi__get16be(s);\r
- return (z << 16) + stbi__get16be(s);\r
-}\r
-\r
-#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF)\r
-// nothing\r
-#else\r
-static int stbi__get16le(stbi__context *s)\r
-{\r
- int z = stbi__get8(s);\r
- return z + (stbi__get8(s) << 8);\r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_BMP\r
-static stbi__uint32 stbi__get32le(stbi__context *s)\r
-{\r
- stbi__uint32 z = stbi__get16le(s);\r
- return z + (stbi__get16le(s) << 16);\r
-}\r
-#endif\r
-\r
-#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings\r
-\r
-\r
-//////////////////////////////////////////////////////////////////////////////\r
-//\r
-// generic converter from built-in img_n to req_comp\r
-// individual types do this automatically as much as possible (e.g. jpeg\r
-// does all cases internally since it needs to colorspace convert anyway,\r
-// and it never has alpha, so very few cases ). png can automatically\r
-// interleave an alpha=255 channel, but falls back to this for other cases\r
-//\r
-// assume data buffer is malloced, so malloc a new one and free that one\r
-// only failure mode is malloc failing\r
-\r
-static stbi_uc stbi__compute_y(int r, int g, int b)\r
-{\r
- return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8);\r
-}\r
-\r
-static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y)\r
-{\r
- int i,j;\r
- unsigned char *good;\r
-\r
- if (req_comp == img_n) return data;\r
- STBI_ASSERT(req_comp >= 1 && req_comp <= 4);\r
-\r
- good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0);\r
- if (good == NULL) {\r
- STBI_FREE(data);\r
- return stbi__errpuc("outofmem", "Out of memory");\r
- }\r
-\r
- for (j=0; j < (int) y; ++j) {\r
- unsigned char *src = data + j * x * img_n ;\r
- unsigned char *dest = good + j * x * req_comp;\r
-\r
- #define STBI__COMBO(a,b) ((a)*8+(b))\r
- #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)\r
- // convert source image with img_n components to one with req_comp components;\r
- // avoid switch per pixel, so use switch per scanline and massive macros\r
- switch (STBI__COMBO(img_n, req_comp)) {\r
- STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break;\r
- STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;\r
- STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break;\r
- STBI__CASE(2,1) { dest[0]=src[0]; } break;\r
- STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;\r
- STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;\r
- STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break;\r
- STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;\r
- STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break;\r
- STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;\r
- STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break;\r
- STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;\r
- default: STBI_ASSERT(0);\r
- }\r
- #undef STBI__CASE\r
- }\r
-\r
- STBI_FREE(data);\r
- return good;\r
-}\r
-\r
-static stbi__uint16 stbi__compute_y_16(int r, int g, int b)\r
-{\r
- return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8);\r
-}\r
-\r
-static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y)\r
-{\r
- int i,j;\r
- stbi__uint16 *good;\r
-\r
- if (req_comp == img_n) return data;\r
- STBI_ASSERT(req_comp >= 1 && req_comp <= 4);\r
-\r
- good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2);\r
- if (good == NULL) {\r
- STBI_FREE(data);\r
- return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");\r
- }\r
-\r
- for (j=0; j < (int) y; ++j) {\r
- stbi__uint16 *src = data + j * x * img_n ;\r
- stbi__uint16 *dest = good + j * x * req_comp;\r
-\r
- #define STBI__COMBO(a,b) ((a)*8+(b))\r
- #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)\r
- // convert source image with img_n components to one with req_comp components;\r
- // avoid switch per pixel, so use switch per scanline and massive macros\r
- switch (STBI__COMBO(img_n, req_comp)) {\r
- STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break;\r
- STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;\r
- STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break;\r
- STBI__CASE(2,1) { dest[0]=src[0]; } break;\r
- STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;\r
- STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;\r
- STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break;\r
- STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;\r
- STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break;\r
- STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;\r
- STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break;\r
- STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;\r
- default: STBI_ASSERT(0);\r
- }\r
- #undef STBI__CASE\r
- }\r
-\r
- STBI_FREE(data);\r
- return good;\r
-}\r
-\r
-#ifndef STBI_NO_LINEAR\r
-static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp)\r
-{\r
- int i,k,n;\r
- float *output;\r
- if (!data) return NULL;\r
- output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0);\r
- if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); }\r
- // compute number of non-alpha components\r
- if (comp & 1) n = comp; else n = comp-1;\r
- for (i=0; i < x*y; ++i) {\r
- for (k=0; k < n; ++k) {\r
- output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale);\r
- }\r
- }\r
- if (n < comp) {\r
- for (i=0; i < x*y; ++i) {\r
- output[i*comp + n] = data[i*comp + n]/255.0f;\r
- }\r
- }\r
- STBI_FREE(data);\r
- return output;\r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_HDR\r
-#define stbi__float2int(x) ((int) (x))\r
-static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp)\r
-{\r
- int i,k,n;\r
- stbi_uc *output;\r
- if (!data) return NULL;\r
- output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0);\r
- if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); }\r
- // compute number of non-alpha components\r
- if (comp & 1) n = comp; else n = comp-1;\r
- for (i=0; i < x*y; ++i) {\r
- for (k=0; k < n; ++k) {\r
- float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f;\r
- if (z < 0) z = 0;\r
- if (z > 255) z = 255;\r
- output[i*comp + k] = (stbi_uc) stbi__float2int(z);\r
- }\r
- if (k < comp) {\r
- float z = data[i*comp+k] * 255 + 0.5f;\r
- if (z < 0) z = 0;\r
- if (z > 255) z = 255;\r
- output[i*comp + k] = (stbi_uc) stbi__float2int(z);\r
- }\r
- }\r
- STBI_FREE(data);\r
- return output;\r
-}\r
-#endif\r
-\r
-//////////////////////////////////////////////////////////////////////////////\r
-//\r
-// "baseline" JPEG/JFIF decoder\r
-//\r
-// simple implementation\r
-// - doesn't support delayed output of y-dimension\r
-// - simple interface (only one output format: 8-bit interleaved RGB)\r
-// - doesn't try to recover corrupt jpegs\r
-// - doesn't allow partial loading, loading multiple at once\r
-// - still fast on x86 (copying globals into locals doesn't help x86)\r
-// - allocates lots of intermediate memory (full size of all components)\r
-// - non-interleaved case requires this anyway\r
-// - allows good upsampling (see next)\r
-// high-quality\r
-// - upsampled channels are bilinearly interpolated, even across blocks\r
-// - quality integer IDCT derived from IJG's 'slow'\r
-// performance\r
-// - fast huffman; reasonable integer IDCT\r
-// - some SIMD kernels for common paths on targets with SSE2/NEON\r
-// - uses a lot of intermediate memory, could cache poorly\r
-\r
-#ifndef STBI_NO_JPEG\r
-\r
-// huffman decoding acceleration\r
-#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache\r
-\r
-typedef struct\r
-{\r
- stbi_uc fast[1 << FAST_BITS];\r
- // weirdly, repacking this into AoS is a 10% speed loss, instead of a win\r
- stbi__uint16 code[256];\r
- stbi_uc values[256];\r
- stbi_uc size[257];\r
- unsigned int maxcode[18];\r
- int delta[17]; // old 'firstsymbol' - old 'firstcode'\r
-} stbi__huffman;\r
-\r
-typedef struct\r
-{\r
- stbi__context *s;\r
- stbi__huffman huff_dc[4];\r
- stbi__huffman huff_ac[4];\r
- stbi__uint16 dequant[4][64];\r
- stbi__int16 fast_ac[4][1 << FAST_BITS];\r
-\r
-// sizes for components, interleaved MCUs\r
- int img_h_max, img_v_max;\r
- int img_mcu_x, img_mcu_y;\r
- int img_mcu_w, img_mcu_h;\r
-\r
-// definition of jpeg image component\r
- struct\r
- {\r
- int id;\r
- int h,v;\r
- int tq;\r
- int hd,ha;\r
- int dc_pred;\r
-\r
- int x,y,w2,h2;\r
- stbi_uc *data;\r
- void *raw_data, *raw_coeff;\r
- stbi_uc *linebuf;\r
- short *coeff; // progressive only\r
- int coeff_w, coeff_h; // number of 8x8 coefficient blocks\r
- } img_comp[4];\r
-\r
- stbi__uint32 code_buffer; // jpeg entropy-coded buffer\r
- int code_bits; // number of valid bits\r
- unsigned char marker; // marker seen while filling entropy buffer\r
- int nomore; // flag if we saw a marker so must stop\r
-\r
- int progressive;\r
- int spec_start;\r
- int spec_end;\r
- int succ_high;\r
- int succ_low;\r
- int eob_run;\r
- int jfif;\r
- int app14_color_transform; // Adobe APP14 tag\r
- int rgb;\r
-\r
- int scan_n, order[4];\r
- int restart_interval, todo;\r
-\r
-// kernels\r
- void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]);\r
- void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step);\r
- stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs);\r
-} stbi__jpeg;\r
-\r
-static int stbi__build_huffman(stbi__huffman *h, int *count)\r
-{\r
- int i,j,k=0;\r
- unsigned int code;\r
- // build size list for each symbol (from JPEG spec)\r
- for (i=0; i < 16; ++i)\r
- for (j=0; j < count[i]; ++j)\r
- h->size[k++] = (stbi_uc) (i+1);\r
- h->size[k] = 0;\r
-\r
- // compute actual symbols (from jpeg spec)\r
- code = 0;\r
- k = 0;\r
- for(j=1; j <= 16; ++j) {\r
- // compute delta to add to code to compute symbol id\r
- h->delta[j] = k - code;\r
- if (h->size[k] == j) {\r
- while (h->size[k] == j)\r
- h->code[k++] = (stbi__uint16) (code++);\r
- if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG");\r
- }\r
- // compute largest code + 1 for this size, preshifted as needed later\r
- h->maxcode[j] = code << (16-j);\r
- code <<= 1;\r
- }\r
- h->maxcode[j] = 0xffffffff;\r
-\r
- // build non-spec acceleration table; 255 is flag for not-accelerated\r
- memset(h->fast, 255, 1 << FAST_BITS);\r
- for (i=0; i < k; ++i) {\r
- int s = h->size[i];\r
- if (s <= FAST_BITS) {\r
- int c = h->code[i] << (FAST_BITS-s);\r
- int m = 1 << (FAST_BITS-s);\r
- for (j=0; j < m; ++j) {\r
- h->fast[c+j] = (stbi_uc) i;\r
- }\r
- }\r
- }\r
- return 1;\r
-}\r
-\r
-// build a table that decodes both magnitude and value of small ACs in\r
-// one go.\r
-static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h)\r
-{\r
- int i;\r
- for (i=0; i < (1 << FAST_BITS); ++i) {\r
- stbi_uc fast = h->fast[i];\r
- fast_ac[i] = 0;\r
- if (fast < 255) {\r
- int rs = h->values[fast];\r
- int run = (rs >> 4) & 15;\r
- int magbits = rs & 15;\r
- int len = h->size[fast];\r
-\r
- if (magbits && len + magbits <= FAST_BITS) {\r
- // magnitude code followed by receive_extend code\r
- int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits);\r
- int m = 1 << (magbits - 1);\r
- if (k < m) k += (~0U << magbits) + 1;\r
- // if the result is small enough, we can fit it in fast_ac table\r
- if (k >= -128 && k <= 127)\r
- fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits));\r
- }\r
- }\r
- }\r
-}\r
-\r
-static void stbi__grow_buffer_unsafe(stbi__jpeg *j)\r
-{\r
- do {\r
- unsigned int b = j->nomore ? 0 : stbi__get8(j->s);\r
- if (b == 0xff) {\r
- int c = stbi__get8(j->s);\r
- while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes\r
- if (c != 0) {\r
- j->marker = (unsigned char) c;\r
- j->nomore = 1;\r
- return;\r
- }\r
- }\r
- j->code_buffer |= b << (24 - j->code_bits);\r
- j->code_bits += 8;\r
- } while (j->code_bits <= 24);\r
-}\r
-\r
-// (1 << n) - 1\r
-static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535};\r
-\r
-// decode a jpeg huffman value from the bitstream\r
-stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h)\r
-{\r
- unsigned int temp;\r
- int c,k;\r
-\r
- if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\r
-\r
- // look at the top FAST_BITS and determine what symbol ID it is,\r
- // if the code is <= FAST_BITS\r
- c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\r
- k = h->fast[c];\r
- if (k < 255) {\r
- int s = h->size[k];\r
- if (s > j->code_bits)\r
- return -1;\r
- j->code_buffer <<= s;\r
- j->code_bits -= s;\r
- return h->values[k];\r
- }\r
-\r
- // naive test is to shift the code_buffer down so k bits are\r
- // valid, then test against maxcode. To speed this up, we've\r
- // preshifted maxcode left so that it has (16-k) 0s at the\r
- // end; in other words, regardless of the number of bits, it\r
- // wants to be compared against something shifted to have 16;\r
- // that way we don't need to shift inside the loop.\r
- temp = j->code_buffer >> 16;\r
- for (k=FAST_BITS+1 ; ; ++k)\r
- if (temp < h->maxcode[k])\r
- break;\r
- if (k == 17) {\r
- // error! code not found\r
- j->code_bits -= 16;\r
- return -1;\r
- }\r
-\r
- if (k > j->code_bits)\r
- return -1;\r
-\r
- // convert the huffman code to the symbol id\r
- c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k];\r
- STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]);\r
-\r
- // convert the id to a symbol\r
- j->code_bits -= k;\r
- j->code_buffer <<= k;\r
- return h->values[c];\r
-}\r
-\r
-// bias[n] = (-1<<n) + 1\r
-static const int stbi__jbias[16] = {0,-1,-3,-7,-15,-31,-63,-127,-255,-511,-1023,-2047,-4095,-8191,-16383,-32767};\r
-\r
-// combined JPEG 'receive' and JPEG 'extend', since baseline\r
-// always extends everything it receives.\r
-stbi_inline static int stbi__extend_receive(stbi__jpeg *j, int n)\r
-{\r
- unsigned int k;\r
- int sgn;\r
- if (j->code_bits < n) stbi__grow_buffer_unsafe(j);\r
-\r
- sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB\r
- k = stbi_lrot(j->code_buffer, n);\r
- STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask)));\r
- j->code_buffer = k & ~stbi__bmask[n];\r
- k &= stbi__bmask[n];\r
- j->code_bits -= n;\r
- return k + (stbi__jbias[n] & ~sgn);\r
-}\r
-\r
-// get some unsigned bits\r
-stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n)\r
-{\r
- unsigned int k;\r
- if (j->code_bits < n) stbi__grow_buffer_unsafe(j);\r
- k = stbi_lrot(j->code_buffer, n);\r
- j->code_buffer = k & ~stbi__bmask[n];\r
- k &= stbi__bmask[n];\r
- j->code_bits -= n;\r
- return k;\r
-}\r
-\r
-stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j)\r
-{\r
- unsigned int k;\r
- if (j->code_bits < 1) stbi__grow_buffer_unsafe(j);\r
- k = j->code_buffer;\r
- j->code_buffer <<= 1;\r
- --j->code_bits;\r
- return k & 0x80000000;\r
-}\r
-\r
-// given a value that's at position X in the zigzag stream,\r
-// where does it appear in the 8x8 matrix coded as row-major?\r
-static const stbi_uc stbi__jpeg_dezigzag[64+15] =\r
-{\r
- 0, 1, 8, 16, 9, 2, 3, 10,\r
- 17, 24, 32, 25, 18, 11, 4, 5,\r
- 12, 19, 26, 33, 40, 48, 41, 34,\r
- 27, 20, 13, 6, 7, 14, 21, 28,\r
- 35, 42, 49, 56, 57, 50, 43, 36,\r
- 29, 22, 15, 23, 30, 37, 44, 51,\r
- 58, 59, 52, 45, 38, 31, 39, 46,\r
- 53, 60, 61, 54, 47, 55, 62, 63,\r
- // let corrupt input sample past end\r
- 63, 63, 63, 63, 63, 63, 63, 63,\r
- 63, 63, 63, 63, 63, 63, 63\r
-};\r
-\r
-// decode one 64-entry block--\r
-static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant)\r
-{\r
- int diff,dc,k;\r
- int t;\r
-\r
- if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\r
- t = stbi__jpeg_huff_decode(j, hdc);\r
- if (t < 0) return stbi__err("bad huffman code","Corrupt JPEG");\r
-\r
- // 0 all the ac values now so we can do it 32-bits at a time\r
- memset(data,0,64*sizeof(data[0]));\r
-\r
- diff = t ? stbi__extend_receive(j, t) : 0;\r
- dc = j->img_comp[b].dc_pred + diff;\r
- j->img_comp[b].dc_pred = dc;\r
- data[0] = (short) (dc * dequant[0]);\r
-\r
- // decode AC components, see JPEG spec\r
- k = 1;\r
- do {\r
- unsigned int zig;\r
- int c,r,s;\r
- if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\r
- c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\r
- r = fac[c];\r
- if (r) { // fast-AC path\r
- k += (r >> 4) & 15; // run\r
- s = r & 15; // combined length\r
- j->code_buffer <<= s;\r
- j->code_bits -= s;\r
- // decode into unzigzag'd location\r
- zig = stbi__jpeg_dezigzag[k++];\r
- data[zig] = (short) ((r >> 8) * dequant[zig]);\r
- } else {\r
- int rs = stbi__jpeg_huff_decode(j, hac);\r
- if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");\r
- s = rs & 15;\r
- r = rs >> 4;\r
- if (s == 0) {\r
- if (rs != 0xf0) break; // end block\r
- k += 16;\r
- } else {\r
- k += r;\r
- // decode into unzigzag'd location\r
- zig = stbi__jpeg_dezigzag[k++];\r
- data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]);\r
- }\r
- }\r
- } while (k < 64);\r
- return 1;\r
-}\r
-\r
-static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b)\r
-{\r
- int diff,dc;\r
- int t;\r
- if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");\r
-\r
- if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\r
-\r
- if (j->succ_high == 0) {\r
- // first scan for DC coefficient, must be first\r
- memset(data,0,64*sizeof(data[0])); // 0 all the ac values now\r
- t = stbi__jpeg_huff_decode(j, hdc);\r
- diff = t ? stbi__extend_receive(j, t) : 0;\r
-\r
- dc = j->img_comp[b].dc_pred + diff;\r
- j->img_comp[b].dc_pred = dc;\r
- data[0] = (short) (dc << j->succ_low);\r
- } else {\r
- // refinement scan for DC coefficient\r
- if (stbi__jpeg_get_bit(j))\r
- data[0] += (short) (1 << j->succ_low);\r
- }\r
- return 1;\r
-}\r
-\r
-// @OPTIMIZE: store non-zigzagged during the decode passes,\r
-// and only de-zigzag when dequantizing\r
-static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac)\r
-{\r
- int k;\r
- if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");\r
-\r
- if (j->succ_high == 0) {\r
- int shift = j->succ_low;\r
-\r
- if (j->eob_run) {\r
- --j->eob_run;\r
- return 1;\r
- }\r
-\r
- k = j->spec_start;\r
- do {\r
- unsigned int zig;\r
- int c,r,s;\r
- if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\r
- c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\r
- r = fac[c];\r
- if (r) { // fast-AC path\r
- k += (r >> 4) & 15; // run\r
- s = r & 15; // combined length\r
- j->code_buffer <<= s;\r
- j->code_bits -= s;\r
- zig = stbi__jpeg_dezigzag[k++];\r
- data[zig] = (short) ((r >> 8) << shift);\r
- } else {\r
- int rs = stbi__jpeg_huff_decode(j, hac);\r
- if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");\r
- s = rs & 15;\r
- r = rs >> 4;\r
- if (s == 0) {\r
- if (r < 15) {\r
- j->eob_run = (1 << r);\r
- if (r)\r
- j->eob_run += stbi__jpeg_get_bits(j, r);\r
- --j->eob_run;\r
- break;\r
- }\r
- k += 16;\r
- } else {\r
- k += r;\r
- zig = stbi__jpeg_dezigzag[k++];\r
- data[zig] = (short) (stbi__extend_receive(j,s) << shift);\r
- }\r
- }\r
- } while (k <= j->spec_end);\r
- } else {\r
- // refinement scan for these AC coefficients\r
-\r
- short bit = (short) (1 << j->succ_low);\r
-\r
- if (j->eob_run) {\r
- --j->eob_run;\r
- for (k = j->spec_start; k <= j->spec_end; ++k) {\r
- short *p = &data[stbi__jpeg_dezigzag[k]];\r
- if (*p != 0)\r
- if (stbi__jpeg_get_bit(j))\r
- if ((*p & bit)==0) {\r
- if (*p > 0)\r
- *p += bit;\r
- else\r
- *p -= bit;\r
- }\r
- }\r
- } else {\r
- k = j->spec_start;\r
- do {\r
- int r,s;\r
- int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh\r
- if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");\r
- s = rs & 15;\r
- r = rs >> 4;\r
- if (s == 0) {\r
- if (r < 15) {\r
- j->eob_run = (1 << r) - 1;\r
- if (r)\r
- j->eob_run += stbi__jpeg_get_bits(j, r);\r
- r = 64; // force end of block\r
- } else {\r
- // r=15 s=0 should write 16 0s, so we just do\r
- // a run of 15 0s and then write s (which is 0),\r
- // so we don't have to do anything special here\r
- }\r
- } else {\r
- if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG");\r
- // sign bit\r
- if (stbi__jpeg_get_bit(j))\r
- s = bit;\r
- else\r
- s = -bit;\r
- }\r
-\r
- // advance by r\r
- while (k <= j->spec_end) {\r
- short *p = &data[stbi__jpeg_dezigzag[k++]];\r
- if (*p != 0) {\r
- if (stbi__jpeg_get_bit(j))\r
- if ((*p & bit)==0) {\r
- if (*p > 0)\r
- *p += bit;\r
- else\r
- *p -= bit;\r
- }\r
- } else {\r
- if (r == 0) {\r
- *p = (short) s;\r
- break;\r
- }\r
- --r;\r
- }\r
- }\r
- } while (k <= j->spec_end);\r
- }\r
- }\r
- return 1;\r
-}\r
-\r
-// take a -128..127 value and stbi__clamp it and convert to 0..255\r
-stbi_inline static stbi_uc stbi__clamp(int x)\r
-{\r
- // trick to use a single test to catch both cases\r
- if ((unsigned int) x > 255) {\r
- if (x < 0) return 0;\r
- if (x > 255) return 255;\r
- }\r
- return (stbi_uc) x;\r
-}\r
-\r
-#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5)))\r
-#define stbi__fsh(x) ((x) * 4096)\r
-\r
-// derived from jidctint -- DCT_ISLOW\r
-#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \\r
- int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \\r
- p2 = s2; \\r
- p3 = s6; \\r
- p1 = (p2+p3) * stbi__f2f(0.5411961f); \\r
- t2 = p1 + p3*stbi__f2f(-1.847759065f); \\r
- t3 = p1 + p2*stbi__f2f( 0.765366865f); \\r
- p2 = s0; \\r
- p3 = s4; \\r
- t0 = stbi__fsh(p2+p3); \\r
- t1 = stbi__fsh(p2-p3); \\r
- x0 = t0+t3; \\r
- x3 = t0-t3; \\r
- x1 = t1+t2; \\r
- x2 = t1-t2; \\r
- t0 = s7; \\r
- t1 = s5; \\r
- t2 = s3; \\r
- t3 = s1; \\r
- p3 = t0+t2; \\r
- p4 = t1+t3; \\r
- p1 = t0+t3; \\r
- p2 = t1+t2; \\r
- p5 = (p3+p4)*stbi__f2f( 1.175875602f); \\r
- t0 = t0*stbi__f2f( 0.298631336f); \\r
- t1 = t1*stbi__f2f( 2.053119869f); \\r
- t2 = t2*stbi__f2f( 3.072711026f); \\r
- t3 = t3*stbi__f2f( 1.501321110f); \\r
- p1 = p5 + p1*stbi__f2f(-0.899976223f); \\r
- p2 = p5 + p2*stbi__f2f(-2.562915447f); \\r
- p3 = p3*stbi__f2f(-1.961570560f); \\r
- p4 = p4*stbi__f2f(-0.390180644f); \\r
- t3 += p1+p4; \\r
- t2 += p2+p3; \\r
- t1 += p2+p4; \\r
- t0 += p1+p3;\r
-\r
-static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64])\r
-{\r
- int i,val[64],*v=val;\r
- stbi_uc *o;\r
- short *d = data;\r
-\r
- // columns\r
- for (i=0; i < 8; ++i,++d, ++v) {\r
- // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing\r
- if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0\r
- && d[40]==0 && d[48]==0 && d[56]==0) {\r
- // no shortcut 0 seconds\r
- // (1|2|3|4|5|6|7)==0 0 seconds\r
- // all separate -0.047 seconds\r
- // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds\r
- int dcterm = d[0]*4;\r
- v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm;\r
- } else {\r
- STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56])\r
- // constants scaled things up by 1<<12; let's bring them back\r
- // down, but keep 2 extra bits of precision\r
- x0 += 512; x1 += 512; x2 += 512; x3 += 512;\r
- v[ 0] = (x0+t3) >> 10;\r
- v[56] = (x0-t3) >> 10;\r
- v[ 8] = (x1+t2) >> 10;\r
- v[48] = (x1-t2) >> 10;\r
- v[16] = (x2+t1) >> 10;\r
- v[40] = (x2-t1) >> 10;\r
- v[24] = (x3+t0) >> 10;\r
- v[32] = (x3-t0) >> 10;\r
- }\r
- }\r
-\r
- for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) {\r
- // no fast case since the first 1D IDCT spread components out\r
- STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7])\r
- // constants scaled things up by 1<<12, plus we had 1<<2 from first\r
- // loop, plus horizontal and vertical each scale by sqrt(8) so together\r
- // we've got an extra 1<<3, so 1<<17 total we need to remove.\r
- // so we want to round that, which means adding 0.5 * 1<<17,\r
- // aka 65536. Also, we'll end up with -128 to 127 that we want\r
- // to encode as 0..255 by adding 128, so we'll add that before the shift\r
- x0 += 65536 + (128<<17);\r
- x1 += 65536 + (128<<17);\r
- x2 += 65536 + (128<<17);\r
- x3 += 65536 + (128<<17);\r
- // tried computing the shifts into temps, or'ing the temps to see\r
- // if any were out of range, but that was slower\r
- o[0] = stbi__clamp((x0+t3) >> 17);\r
- o[7] = stbi__clamp((x0-t3) >> 17);\r
- o[1] = stbi__clamp((x1+t2) >> 17);\r
- o[6] = stbi__clamp((x1-t2) >> 17);\r
- o[2] = stbi__clamp((x2+t1) >> 17);\r
- o[5] = stbi__clamp((x2-t1) >> 17);\r
- o[3] = stbi__clamp((x3+t0) >> 17);\r
- o[4] = stbi__clamp((x3-t0) >> 17);\r
- }\r
-}\r
-\r
-#ifdef STBI_SSE2\r
-// sse2 integer IDCT. not the fastest possible implementation but it\r
-// produces bit-identical results to the generic C version so it's\r
-// fully "transparent".\r
-static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])\r
-{\r
- // This is constructed to match our regular (generic) integer IDCT exactly.\r
- __m128i row0, row1, row2, row3, row4, row5, row6, row7;\r
- __m128i tmp;\r
-\r
- // dot product constant: even elems=x, odd elems=y\r
- #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y))\r
-\r
- // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit)\r
- // out(1) = c1[even]*x + c1[odd]*y\r
- #define dct_rot(out0,out1, x,y,c0,c1) \\r
- __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \\r
- __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \\r
- __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \\r
- __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \\r
- __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \\r
- __m128i out1##_h = _mm_madd_epi16(c0##hi, c1)\r
-\r
- // out = in << 12 (in 16-bit, out 32-bit)\r
- #define dct_widen(out, in) \\r
- __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \\r
- __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4)\r
-\r
- // wide add\r
- #define dct_wadd(out, a, b) \\r
- __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \\r
- __m128i out##_h = _mm_add_epi32(a##_h, b##_h)\r
-\r
- // wide sub\r
- #define dct_wsub(out, a, b) \\r
- __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \\r
- __m128i out##_h = _mm_sub_epi32(a##_h, b##_h)\r
-\r
- // butterfly a/b, add bias, then shift by "s" and pack\r
- #define dct_bfly32o(out0, out1, a,b,bias,s) \\r
- { \\r
- __m128i abiased_l = _mm_add_epi32(a##_l, bias); \\r
- __m128i abiased_h = _mm_add_epi32(a##_h, bias); \\r
- dct_wadd(sum, abiased, b); \\r
- dct_wsub(dif, abiased, b); \\r
- out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \\r
- out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \\r
- }\r
-\r
- // 8-bit interleave step (for transposes)\r
- #define dct_interleave8(a, b) \\r
- tmp = a; \\r
- a = _mm_unpacklo_epi8(a, b); \\r
- b = _mm_unpackhi_epi8(tmp, b)\r
-\r
- // 16-bit interleave step (for transposes)\r
- #define dct_interleave16(a, b) \\r
- tmp = a; \\r
- a = _mm_unpacklo_epi16(a, b); \\r
- b = _mm_unpackhi_epi16(tmp, b)\r
-\r
- #define dct_pass(bias,shift) \\r
- { \\r
- /* even part */ \\r
- dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \\r
- __m128i sum04 = _mm_add_epi16(row0, row4); \\r
- __m128i dif04 = _mm_sub_epi16(row0, row4); \\r
- dct_widen(t0e, sum04); \\r
- dct_widen(t1e, dif04); \\r
- dct_wadd(x0, t0e, t3e); \\r
- dct_wsub(x3, t0e, t3e); \\r
- dct_wadd(x1, t1e, t2e); \\r
- dct_wsub(x2, t1e, t2e); \\r
- /* odd part */ \\r
- dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \\r
- dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \\r
- __m128i sum17 = _mm_add_epi16(row1, row7); \\r
- __m128i sum35 = _mm_add_epi16(row3, row5); \\r
- dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \\r
- dct_wadd(x4, y0o, y4o); \\r
- dct_wadd(x5, y1o, y5o); \\r
- dct_wadd(x6, y2o, y5o); \\r
- dct_wadd(x7, y3o, y4o); \\r
- dct_bfly32o(row0,row7, x0,x7,bias,shift); \\r
- dct_bfly32o(row1,row6, x1,x6,bias,shift); \\r
- dct_bfly32o(row2,row5, x2,x5,bias,shift); \\r
- dct_bfly32o(row3,row4, x3,x4,bias,shift); \\r
- }\r
-\r
- __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f));\r
- __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f));\r
- __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f));\r
- __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f));\r
- __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f));\r
- __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f));\r
- __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f));\r
- __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f));\r
-\r
- // rounding biases in column/row passes, see stbi__idct_block for explanation.\r
- __m128i bias_0 = _mm_set1_epi32(512);\r
- __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17));\r
-\r
- // load\r
- row0 = _mm_load_si128((const __m128i *) (data + 0*8));\r
- row1 = _mm_load_si128((const __m128i *) (data + 1*8));\r
- row2 = _mm_load_si128((const __m128i *) (data + 2*8));\r
- row3 = _mm_load_si128((const __m128i *) (data + 3*8));\r
- row4 = _mm_load_si128((const __m128i *) (data + 4*8));\r
- row5 = _mm_load_si128((const __m128i *) (data + 5*8));\r
- row6 = _mm_load_si128((const __m128i *) (data + 6*8));\r
- row7 = _mm_load_si128((const __m128i *) (data + 7*8));\r
-\r
- // column pass\r
- dct_pass(bias_0, 10);\r
-\r
- {\r
- // 16bit 8x8 transpose pass 1\r
- dct_interleave16(row0, row4);\r
- dct_interleave16(row1, row5);\r
- dct_interleave16(row2, row6);\r
- dct_interleave16(row3, row7);\r
-\r
- // transpose pass 2\r
- dct_interleave16(row0, row2);\r
- dct_interleave16(row1, row3);\r
- dct_interleave16(row4, row6);\r
- dct_interleave16(row5, row7);\r
-\r
- // transpose pass 3\r
- dct_interleave16(row0, row1);\r
- dct_interleave16(row2, row3);\r
- dct_interleave16(row4, row5);\r
- dct_interleave16(row6, row7);\r
- }\r
-\r
- // row pass\r
- dct_pass(bias_1, 17);\r
-\r
- {\r
- // pack\r
- __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7\r
- __m128i p1 = _mm_packus_epi16(row2, row3);\r
- __m128i p2 = _mm_packus_epi16(row4, row5);\r
- __m128i p3 = _mm_packus_epi16(row6, row7);\r
-\r
- // 8bit 8x8 transpose pass 1\r
- dct_interleave8(p0, p2); // a0e0a1e1...\r
- dct_interleave8(p1, p3); // c0g0c1g1...\r
-\r
- // transpose pass 2\r
- dct_interleave8(p0, p1); // a0c0e0g0...\r
- dct_interleave8(p2, p3); // b0d0f0h0...\r
-\r
- // transpose pass 3\r
- dct_interleave8(p0, p2); // a0b0c0d0...\r
- dct_interleave8(p1, p3); // a4b4c4d4...\r
-\r
- // store\r
- _mm_storel_epi64((__m128i *) out, p0); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, p2); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, p1); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, p3); out += out_stride;\r
- _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e));\r
- }\r
-\r
-#undef dct_const\r
-#undef dct_rot\r
-#undef dct_widen\r
-#undef dct_wadd\r
-#undef dct_wsub\r
-#undef dct_bfly32o\r
-#undef dct_interleave8\r
-#undef dct_interleave16\r
-#undef dct_pass\r
-}\r
-\r
-#endif // STBI_SSE2\r
-\r
-#ifdef STBI_NEON\r
-\r
-// NEON integer IDCT. should produce bit-identical\r
-// results to the generic C version.\r
-static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])\r
-{\r
- int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;\r
-\r
- int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f));\r
- int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f));\r
- int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f));\r
- int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f));\r
- int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f));\r
- int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f));\r
- int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f));\r
- int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f));\r
- int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f));\r
- int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f));\r
- int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f));\r
- int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f));\r
-\r
-#define dct_long_mul(out, inq, coeff) \\r
- int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \\r
- int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff)\r
-\r
-#define dct_long_mac(out, acc, inq, coeff) \\r
- int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \\r
- int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff)\r
-\r
-#define dct_widen(out, inq) \\r
- int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \\r
- int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12)\r
-\r
-// wide add\r
-#define dct_wadd(out, a, b) \\r
- int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \\r
- int32x4_t out##_h = vaddq_s32(a##_h, b##_h)\r
-\r
-// wide sub\r
-#define dct_wsub(out, a, b) \\r
- int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \\r
- int32x4_t out##_h = vsubq_s32(a##_h, b##_h)\r
-\r
-// butterfly a/b, then shift using "shiftop" by "s" and pack\r
-#define dct_bfly32o(out0,out1, a,b,shiftop,s) \\r
- { \\r
- dct_wadd(sum, a, b); \\r
- dct_wsub(dif, a, b); \\r
- out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \\r
- out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \\r
- }\r
-\r
-#define dct_pass(shiftop, shift) \\r
- { \\r
- /* even part */ \\r
- int16x8_t sum26 = vaddq_s16(row2, row6); \\r
- dct_long_mul(p1e, sum26, rot0_0); \\r
- dct_long_mac(t2e, p1e, row6, rot0_1); \\r
- dct_long_mac(t3e, p1e, row2, rot0_2); \\r
- int16x8_t sum04 = vaddq_s16(row0, row4); \\r
- int16x8_t dif04 = vsubq_s16(row0, row4); \\r
- dct_widen(t0e, sum04); \\r
- dct_widen(t1e, dif04); \\r
- dct_wadd(x0, t0e, t3e); \\r
- dct_wsub(x3, t0e, t3e); \\r
- dct_wadd(x1, t1e, t2e); \\r
- dct_wsub(x2, t1e, t2e); \\r
- /* odd part */ \\r
- int16x8_t sum15 = vaddq_s16(row1, row5); \\r
- int16x8_t sum17 = vaddq_s16(row1, row7); \\r
- int16x8_t sum35 = vaddq_s16(row3, row5); \\r
- int16x8_t sum37 = vaddq_s16(row3, row7); \\r
- int16x8_t sumodd = vaddq_s16(sum17, sum35); \\r
- dct_long_mul(p5o, sumodd, rot1_0); \\r
- dct_long_mac(p1o, p5o, sum17, rot1_1); \\r
- dct_long_mac(p2o, p5o, sum35, rot1_2); \\r
- dct_long_mul(p3o, sum37, rot2_0); \\r
- dct_long_mul(p4o, sum15, rot2_1); \\r
- dct_wadd(sump13o, p1o, p3o); \\r
- dct_wadd(sump24o, p2o, p4o); \\r
- dct_wadd(sump23o, p2o, p3o); \\r
- dct_wadd(sump14o, p1o, p4o); \\r
- dct_long_mac(x4, sump13o, row7, rot3_0); \\r
- dct_long_mac(x5, sump24o, row5, rot3_1); \\r
- dct_long_mac(x6, sump23o, row3, rot3_2); \\r
- dct_long_mac(x7, sump14o, row1, rot3_3); \\r
- dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \\r
- dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \\r
- dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \\r
- dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \\r
- }\r
-\r
- // load\r
- row0 = vld1q_s16(data + 0*8);\r
- row1 = vld1q_s16(data + 1*8);\r
- row2 = vld1q_s16(data + 2*8);\r
- row3 = vld1q_s16(data + 3*8);\r
- row4 = vld1q_s16(data + 4*8);\r
- row5 = vld1q_s16(data + 5*8);\r
- row6 = vld1q_s16(data + 6*8);\r
- row7 = vld1q_s16(data + 7*8);\r
-\r
- // add DC bias\r
- row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0));\r
-\r
- // column pass\r
- dct_pass(vrshrn_n_s32, 10);\r
-\r
- // 16bit 8x8 transpose\r
- {\r
-// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively.\r
-// whether compilers actually get this is another story, sadly.\r
-#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; }\r
-#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); }\r
-#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); }\r
-\r
- // pass 1\r
- dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6\r
- dct_trn16(row2, row3);\r
- dct_trn16(row4, row5);\r
- dct_trn16(row6, row7);\r
-\r
- // pass 2\r
- dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4\r
- dct_trn32(row1, row3);\r
- dct_trn32(row4, row6);\r
- dct_trn32(row5, row7);\r
-\r
- // pass 3\r
- dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0\r
- dct_trn64(row1, row5);\r
- dct_trn64(row2, row6);\r
- dct_trn64(row3, row7);\r
-\r
-#undef dct_trn16\r
-#undef dct_trn32\r
-#undef dct_trn64\r
- }\r
-\r
- // row pass\r
- // vrshrn_n_s32 only supports shifts up to 16, we need\r
- // 17. so do a non-rounding shift of 16 first then follow\r
- // up with a rounding shift by 1.\r
- dct_pass(vshrn_n_s32, 16);\r
-\r
- {\r
- // pack and round\r
- uint8x8_t p0 = vqrshrun_n_s16(row0, 1);\r
- uint8x8_t p1 = vqrshrun_n_s16(row1, 1);\r
- uint8x8_t p2 = vqrshrun_n_s16(row2, 1);\r
- uint8x8_t p3 = vqrshrun_n_s16(row3, 1);\r
- uint8x8_t p4 = vqrshrun_n_s16(row4, 1);\r
- uint8x8_t p5 = vqrshrun_n_s16(row5, 1);\r
- uint8x8_t p6 = vqrshrun_n_s16(row6, 1);\r
- uint8x8_t p7 = vqrshrun_n_s16(row7, 1);\r
-\r
- // again, these can translate into one instruction, but often don't.\r
-#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; }\r
-#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); }\r
-#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); }\r
-\r
- // sadly can't use interleaved stores here since we only write\r
- // 8 bytes to each scan line!\r
-\r
- // 8x8 8-bit transpose pass 1\r
- dct_trn8_8(p0, p1);\r
- dct_trn8_8(p2, p3);\r
- dct_trn8_8(p4, p5);\r
- dct_trn8_8(p6, p7);\r
-\r
- // pass 2\r
- dct_trn8_16(p0, p2);\r
- dct_trn8_16(p1, p3);\r
- dct_trn8_16(p4, p6);\r
- dct_trn8_16(p5, p7);\r
-\r
- // pass 3\r
- dct_trn8_32(p0, p4);\r
- dct_trn8_32(p1, p5);\r
- dct_trn8_32(p2, p6);\r
- dct_trn8_32(p3, p7);\r
-\r
- // store\r
- vst1_u8(out, p0); out += out_stride;\r
- vst1_u8(out, p1); out += out_stride;\r
- vst1_u8(out, p2); out += out_stride;\r
- vst1_u8(out, p3); out += out_stride;\r
- vst1_u8(out, p4); out += out_stride;\r
- vst1_u8(out, p5); out += out_stride;\r
- vst1_u8(out, p6); out += out_stride;\r
- vst1_u8(out, p7);\r
-\r
-#undef dct_trn8_8\r
-#undef dct_trn8_16\r
-#undef dct_trn8_32\r
- }\r
-\r
-#undef dct_long_mul\r
-#undef dct_long_mac\r
-#undef dct_widen\r
-#undef dct_wadd\r
-#undef dct_wsub\r
-#undef dct_bfly32o\r
-#undef dct_pass\r
-}\r
-\r
-#endif // STBI_NEON\r
-\r
-#define STBI__MARKER_none 0xff\r
-// if there's a pending marker from the entropy stream, return that\r
-// otherwise, fetch from the stream and get a marker. if there's no\r
-// marker, return 0xff, which is never a valid marker value\r
-static stbi_uc stbi__get_marker(stbi__jpeg *j)\r
-{\r
- stbi_uc x;\r
- if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; }\r
- x = stbi__get8(j->s);\r
- if (x != 0xff) return STBI__MARKER_none;\r
- while (x == 0xff)\r
- x = stbi__get8(j->s); // consume repeated 0xff fill bytes\r
- return x;\r
-}\r
-\r
-// in each scan, we'll have scan_n components, and the order\r
-// of the components is specified by order[]\r
-#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7)\r
-\r
-// after a restart interval, stbi__jpeg_reset the entropy decoder and\r
-// the dc prediction\r
-static void stbi__jpeg_reset(stbi__jpeg *j)\r
-{\r
- j->code_bits = 0;\r
- j->code_buffer = 0;\r
- j->nomore = 0;\r
- j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0;\r
- j->marker = STBI__MARKER_none;\r
- j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff;\r
- j->eob_run = 0;\r
- // no more than 1<<31 MCUs if no restart_interal? that's plenty safe,\r
- // since we don't even allow 1<<30 pixels\r
-}\r
-\r
-static int stbi__parse_entropy_coded_data(stbi__jpeg *z)\r
-{\r
- stbi__jpeg_reset(z);\r
- if (!z->progressive) {\r
- if (z->scan_n == 1) {\r
- int i,j;\r
- STBI_SIMD_ALIGN(short, data[64]);\r
- int n = z->order[0];\r
- // non-interleaved data, we just need to process one block at a time,\r
- // in trivial scanline order\r
- // number of blocks to do just depends on how many actual "pixels" this\r
- // component has, independent of interleaved MCU blocking and such\r
- int w = (z->img_comp[n].x+7) >> 3;\r
- int h = (z->img_comp[n].y+7) >> 3;\r
- for (j=0; j < h; ++j) {\r
- for (i=0; i < w; ++i) {\r
- int ha = z->img_comp[n].ha;\r
- if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;\r
- z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);\r
- // every data block is an MCU, so countdown the restart interval\r
- if (--z->todo <= 0) {\r
- if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\r
- // if it's NOT a restart, then just bail, so we get corrupt data\r
- // rather than no data\r
- if (!STBI__RESTART(z->marker)) return 1;\r
- stbi__jpeg_reset(z);\r
- }\r
- }\r
- }\r
- return 1;\r
- } else { // interleaved\r
- int i,j,k,x,y;\r
- STBI_SIMD_ALIGN(short, data[64]);\r
- for (j=0; j < z->img_mcu_y; ++j) {\r
- for (i=0; i < z->img_mcu_x; ++i) {\r
- // scan an interleaved mcu... process scan_n components in order\r
- for (k=0; k < z->scan_n; ++k) {\r
- int n = z->order[k];\r
- // scan out an mcu's worth of this component; that's just determined\r
- // by the basic H and V specified for the component\r
- for (y=0; y < z->img_comp[n].v; ++y) {\r
- for (x=0; x < z->img_comp[n].h; ++x) {\r
- int x2 = (i*z->img_comp[n].h + x)*8;\r
- int y2 = (j*z->img_comp[n].v + y)*8;\r
- int ha = z->img_comp[n].ha;\r
- if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;\r
- z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data);\r
- }\r
- }\r
- }\r
- // after all interleaved components, that's an interleaved MCU,\r
- // so now count down the restart interval\r
- if (--z->todo <= 0) {\r
- if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\r
- if (!STBI__RESTART(z->marker)) return 1;\r
- stbi__jpeg_reset(z);\r
- }\r
- }\r
- }\r
- return 1;\r
- }\r
- } else {\r
- if (z->scan_n == 1) {\r
- int i,j;\r
- int n = z->order[0];\r
- // non-interleaved data, we just need to process one block at a time,\r
- // in trivial scanline order\r
- // number of blocks to do just depends on how many actual "pixels" this\r
- // component has, independent of interleaved MCU blocking and such\r
- int w = (z->img_comp[n].x+7) >> 3;\r
- int h = (z->img_comp[n].y+7) >> 3;\r
- for (j=0; j < h; ++j) {\r
- for (i=0; i < w; ++i) {\r
- short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);\r
- if (z->spec_start == 0) {\r
- if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))\r
- return 0;\r
- } else {\r
- int ha = z->img_comp[n].ha;\r
- if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha]))\r
- return 0;\r
- }\r
- // every data block is an MCU, so countdown the restart interval\r
- if (--z->todo <= 0) {\r
- if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\r
- if (!STBI__RESTART(z->marker)) return 1;\r
- stbi__jpeg_reset(z);\r
- }\r
- }\r
- }\r
- return 1;\r
- } else { // interleaved\r
- int i,j,k,x,y;\r
- for (j=0; j < z->img_mcu_y; ++j) {\r
- for (i=0; i < z->img_mcu_x; ++i) {\r
- // scan an interleaved mcu... process scan_n components in order\r
- for (k=0; k < z->scan_n; ++k) {\r
- int n = z->order[k];\r
- // scan out an mcu's worth of this component; that's just determined\r
- // by the basic H and V specified for the component\r
- for (y=0; y < z->img_comp[n].v; ++y) {\r
- for (x=0; x < z->img_comp[n].h; ++x) {\r
- int x2 = (i*z->img_comp[n].h + x);\r
- int y2 = (j*z->img_comp[n].v + y);\r
- short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w);\r
- if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))\r
- return 0;\r
- }\r
- }\r
- }\r
- // after all interleaved components, that's an interleaved MCU,\r
- // so now count down the restart interval\r
- if (--z->todo <= 0) {\r
- if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\r
- if (!STBI__RESTART(z->marker)) return 1;\r
- stbi__jpeg_reset(z);\r
- }\r
- }\r
- }\r
- return 1;\r
- }\r
- }\r
-}\r
-\r
-static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant)\r
-{\r
- int i;\r
- for (i=0; i < 64; ++i)\r
- data[i] *= dequant[i];\r
-}\r
-\r
-static void stbi__jpeg_finish(stbi__jpeg *z)\r
-{\r
- if (z->progressive) {\r
- // dequantize and idct the data\r
- int i,j,n;\r
- for (n=0; n < z->s->img_n; ++n) {\r
- int w = (z->img_comp[n].x+7) >> 3;\r
- int h = (z->img_comp[n].y+7) >> 3;\r
- for (j=0; j < h; ++j) {\r
- for (i=0; i < w; ++i) {\r
- short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);\r
- stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]);\r
- z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);\r
- }\r
- }\r
- }\r
- }\r
-}\r
-\r
-static int stbi__process_marker(stbi__jpeg *z, int m)\r
-{\r
- int L;\r
- switch (m) {\r
- case STBI__MARKER_none: // no marker found\r
- return stbi__err("expected marker","Corrupt JPEG");\r
-\r
- case 0xDD: // DRI - specify restart interval\r
- if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG");\r
- z->restart_interval = stbi__get16be(z->s);\r
- return 1;\r
-\r
- case 0xDB: // DQT - define quantization table\r
- L = stbi__get16be(z->s)-2;\r
- while (L > 0) {\r
- int q = stbi__get8(z->s);\r
- int p = q >> 4, sixteen = (p != 0);\r
- int t = q & 15,i;\r
- if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG");\r
- if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG");\r
-\r
- for (i=0; i < 64; ++i)\r
- z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s));\r
- L -= (sixteen ? 129 : 65);\r
- }\r
- return L==0;\r
-\r
- case 0xC4: // DHT - define huffman table\r
- L = stbi__get16be(z->s)-2;\r
- while (L > 0) {\r
- stbi_uc *v;\r
- int sizes[16],i,n=0;\r
- int q = stbi__get8(z->s);\r
- int tc = q >> 4;\r
- int th = q & 15;\r
- if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG");\r
- for (i=0; i < 16; ++i) {\r
- sizes[i] = stbi__get8(z->s);\r
- n += sizes[i];\r
- }\r
- L -= 17;\r
- if (tc == 0) {\r
- if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0;\r
- v = z->huff_dc[th].values;\r
- } else {\r
- if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0;\r
- v = z->huff_ac[th].values;\r
- }\r
- for (i=0; i < n; ++i)\r
- v[i] = stbi__get8(z->s);\r
- if (tc != 0)\r
- stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th);\r
- L -= n;\r
- }\r
- return L==0;\r
- }\r
-\r
- // check for comment block or APP blocks\r
- if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) {\r
- L = stbi__get16be(z->s);\r
- if (L < 2) {\r
- if (m == 0xFE)\r
- return stbi__err("bad COM len","Corrupt JPEG");\r
- else\r
- return stbi__err("bad APP len","Corrupt JPEG");\r
- }\r
- L -= 2;\r
-\r
- if (m == 0xE0 && L >= 5) { // JFIF APP0 segment\r
- static const unsigned char tag[5] = {'J','F','I','F','\0'};\r
- int ok = 1;\r
- int i;\r
- for (i=0; i < 5; ++i)\r
- if (stbi__get8(z->s) != tag[i])\r
- ok = 0;\r
- L -= 5;\r
- if (ok)\r
- z->jfif = 1;\r
- } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment\r
- static const unsigned char tag[6] = {'A','d','o','b','e','\0'};\r
- int ok = 1;\r
- int i;\r
- for (i=0; i < 6; ++i)\r
- if (stbi__get8(z->s) != tag[i])\r
- ok = 0;\r
- L -= 6;\r
- if (ok) {\r
- stbi__get8(z->s); // version\r
- stbi__get16be(z->s); // flags0\r
- stbi__get16be(z->s); // flags1\r
- z->app14_color_transform = stbi__get8(z->s); // color transform\r
- L -= 6;\r
- }\r
- }\r
-\r
- stbi__skip(z->s, L);\r
- return 1;\r
- }\r
-\r
- return stbi__err("unknown marker","Corrupt JPEG");\r
-}\r
-\r
-// after we see SOS\r
-static int stbi__process_scan_header(stbi__jpeg *z)\r
-{\r
- int i;\r
- int Ls = stbi__get16be(z->s);\r
- z->scan_n = stbi__get8(z->s);\r
- if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG");\r
- if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG");\r
- for (i=0; i < z->scan_n; ++i) {\r
- int id = stbi__get8(z->s), which;\r
- int q = stbi__get8(z->s);\r
- for (which = 0; which < z->s->img_n; ++which)\r
- if (z->img_comp[which].id == id)\r
- break;\r
- if (which == z->s->img_n) return 0; // no match\r
- z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG");\r
- z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG");\r
- z->order[i] = which;\r
- }\r
-\r
- {\r
- int aa;\r
- z->spec_start = stbi__get8(z->s);\r
- z->spec_end = stbi__get8(z->s); // should be 63, but might be 0\r
- aa = stbi__get8(z->s);\r
- z->succ_high = (aa >> 4);\r
- z->succ_low = (aa & 15);\r
- if (z->progressive) {\r
- if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13)\r
- return stbi__err("bad SOS", "Corrupt JPEG");\r
- } else {\r
- if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG");\r
- if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG");\r
- z->spec_end = 63;\r
- }\r
- }\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why)\r
-{\r
- int i;\r
- for (i=0; i < ncomp; ++i) {\r
- if (z->img_comp[i].raw_data) {\r
- STBI_FREE(z->img_comp[i].raw_data);\r
- z->img_comp[i].raw_data = NULL;\r
- z->img_comp[i].data = NULL;\r
- }\r
- if (z->img_comp[i].raw_coeff) {\r
- STBI_FREE(z->img_comp[i].raw_coeff);\r
- z->img_comp[i].raw_coeff = 0;\r
- z->img_comp[i].coeff = 0;\r
- }\r
- if (z->img_comp[i].linebuf) {\r
- STBI_FREE(z->img_comp[i].linebuf);\r
- z->img_comp[i].linebuf = NULL;\r
- }\r
- }\r
- return why;\r
-}\r
-\r
-static int stbi__process_frame_header(stbi__jpeg *z, int scan)\r
-{\r
- stbi__context *s = z->s;\r
- int Lf,p,i,q, h_max=1,v_max=1,c;\r
- Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG\r
- p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline\r
- s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG\r
- s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires\r
- c = stbi__get8(s);\r
- if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG");\r
- s->img_n = c;\r
- for (i=0; i < c; ++i) {\r
- z->img_comp[i].data = NULL;\r
- z->img_comp[i].linebuf = NULL;\r
- }\r
-\r
- if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG");\r
-\r
- z->rgb = 0;\r
- for (i=0; i < s->img_n; ++i) {\r
- static const unsigned char rgb[3] = { 'R', 'G', 'B' };\r
- z->img_comp[i].id = stbi__get8(s);\r
- if (s->img_n == 3 && z->img_comp[i].id == rgb[i])\r
- ++z->rgb;\r
- q = stbi__get8(s);\r
- z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG");\r
- z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG");\r
- z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG");\r
- }\r
-\r
- if (scan != STBI__SCAN_load) return 1;\r
-\r
- if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode");\r
-\r
- for (i=0; i < s->img_n; ++i) {\r
- if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h;\r
- if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v;\r
- }\r
-\r
- // compute interleaved mcu info\r
- z->img_h_max = h_max;\r
- z->img_v_max = v_max;\r
- z->img_mcu_w = h_max * 8;\r
- z->img_mcu_h = v_max * 8;\r
- // these sizes can't be more than 17 bits\r
- z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w;\r
- z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h;\r
-\r
- for (i=0; i < s->img_n; ++i) {\r
- // number of effective pixels (e.g. for non-interleaved MCU)\r
- z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max;\r
- z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max;\r
- // to simplify generation, we'll allocate enough memory to decode\r
- // the bogus oversized data from using interleaved MCUs and their\r
- // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't\r
- // discard the extra data until colorspace conversion\r
- //\r
- // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier)\r
- // so these muls can't overflow with 32-bit ints (which we require)\r
- z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8;\r
- z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8;\r
- z->img_comp[i].coeff = 0;\r
- z->img_comp[i].raw_coeff = 0;\r
- z->img_comp[i].linebuf = NULL;\r
- z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15);\r
- if (z->img_comp[i].raw_data == NULL)\r
- return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));\r
- // align blocks for idct using mmx/sse\r
- z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15);\r
- if (z->progressive) {\r
- // w2, h2 are multiples of 8 (see above)\r
- z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8;\r
- z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8;\r
- z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15);\r
- if (z->img_comp[i].raw_coeff == NULL)\r
- return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));\r
- z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15);\r
- }\r
- }\r
-\r
- return 1;\r
-}\r
-\r
-// use comparisons since in some cases we handle more than one case (e.g. SOF)\r
-#define stbi__DNL(x) ((x) == 0xdc)\r
-#define stbi__SOI(x) ((x) == 0xd8)\r
-#define stbi__EOI(x) ((x) == 0xd9)\r
-#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2)\r
-#define stbi__SOS(x) ((x) == 0xda)\r
-\r
-#define stbi__SOF_progressive(x) ((x) == 0xc2)\r
-\r
-static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan)\r
-{\r
- int m;\r
- z->jfif = 0;\r
- z->app14_color_transform = -1; // valid values are 0,1,2\r
- z->marker = STBI__MARKER_none; // initialize cached marker to empty\r
- m = stbi__get_marker(z);\r
- if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG");\r
- if (scan == STBI__SCAN_type) return 1;\r
- m = stbi__get_marker(z);\r
- while (!stbi__SOF(m)) {\r
- if (!stbi__process_marker(z,m)) return 0;\r
- m = stbi__get_marker(z);\r
- while (m == STBI__MARKER_none) {\r
- // some files have extra padding after their blocks, so ok, we'll scan\r
- if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG");\r
- m = stbi__get_marker(z);\r
- }\r
- }\r
- z->progressive = stbi__SOF_progressive(m);\r
- if (!stbi__process_frame_header(z, scan)) return 0;\r
- return 1;\r
-}\r
-\r
-// decode image to YCbCr format\r
-static int stbi__decode_jpeg_image(stbi__jpeg *j)\r
-{\r
- int m;\r
- for (m = 0; m < 4; m++) {\r
- j->img_comp[m].raw_data = NULL;\r
- j->img_comp[m].raw_coeff = NULL;\r
- }\r
- j->restart_interval = 0;\r
- if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0;\r
- m = stbi__get_marker(j);\r
- while (!stbi__EOI(m)) {\r
- if (stbi__SOS(m)) {\r
- if (!stbi__process_scan_header(j)) return 0;\r
- if (!stbi__parse_entropy_coded_data(j)) return 0;\r
- if (j->marker == STBI__MARKER_none ) {\r
- // handle 0s at the end of image data from IP Kamera 9060\r
- while (!stbi__at_eof(j->s)) {\r
- int x = stbi__get8(j->s);\r
- if (x == 255) {\r
- j->marker = stbi__get8(j->s);\r
- break;\r
- }\r
- }\r
- // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0\r
- }\r
- } else if (stbi__DNL(m)) {\r
- int Ld = stbi__get16be(j->s);\r
- stbi__uint32 NL = stbi__get16be(j->s);\r
- if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG");\r
- if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG");\r
- } else {\r
- if (!stbi__process_marker(j, m)) return 0;\r
- }\r
- m = stbi__get_marker(j);\r
- }\r
- if (j->progressive)\r
- stbi__jpeg_finish(j);\r
- return 1;\r
-}\r
-\r
-// static jfif-centered resampling (across block boundaries)\r
-\r
-typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1,\r
- int w, int hs);\r
-\r
-#define stbi__div4(x) ((stbi_uc) ((x) >> 2))\r
-\r
-static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- STBI_NOTUSED(out);\r
- STBI_NOTUSED(in_far);\r
- STBI_NOTUSED(w);\r
- STBI_NOTUSED(hs);\r
- return in_near;\r
-}\r
-\r
-static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- // need to generate two samples vertically for every one in input\r
- int i;\r
- STBI_NOTUSED(hs);\r
- for (i=0; i < w; ++i)\r
- out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2);\r
- return out;\r
-}\r
-\r
-static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- // need to generate two samples horizontally for every one in input\r
- int i;\r
- stbi_uc *input = in_near;\r
-\r
- if (w == 1) {\r
- // if only one sample, can't do any interpolation\r
- out[0] = out[1] = input[0];\r
- return out;\r
- }\r
-\r
- out[0] = input[0];\r
- out[1] = stbi__div4(input[0]*3 + input[1] + 2);\r
- for (i=1; i < w-1; ++i) {\r
- int n = 3*input[i]+2;\r
- out[i*2+0] = stbi__div4(n+input[i-1]);\r
- out[i*2+1] = stbi__div4(n+input[i+1]);\r
- }\r
- out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2);\r
- out[i*2+1] = input[w-1];\r
-\r
- STBI_NOTUSED(in_far);\r
- STBI_NOTUSED(hs);\r
-\r
- return out;\r
-}\r
-\r
-#define stbi__div16(x) ((stbi_uc) ((x) >> 4))\r
-\r
-static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- // need to generate 2x2 samples for every one in input\r
- int i,t0,t1;\r
- if (w == 1) {\r
- out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);\r
- return out;\r
- }\r
-\r
- t1 = 3*in_near[0] + in_far[0];\r
- out[0] = stbi__div4(t1+2);\r
- for (i=1; i < w; ++i) {\r
- t0 = t1;\r
- t1 = 3*in_near[i]+in_far[i];\r
- out[i*2-1] = stbi__div16(3*t0 + t1 + 8);\r
- out[i*2 ] = stbi__div16(3*t1 + t0 + 8);\r
- }\r
- out[w*2-1] = stbi__div4(t1+2);\r
-\r
- STBI_NOTUSED(hs);\r
-\r
- return out;\r
-}\r
-\r
-#if defined(STBI_SSE2) || defined(STBI_NEON)\r
-static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- // need to generate 2x2 samples for every one in input\r
- int i=0,t0,t1;\r
-\r
- if (w == 1) {\r
- out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);\r
- return out;\r
- }\r
-\r
- t1 = 3*in_near[0] + in_far[0];\r
- // process groups of 8 pixels for as long as we can.\r
- // note we can't handle the last pixel in a row in this loop\r
- // because we need to handle the filter boundary conditions.\r
- for (; i < ((w-1) & ~7); i += 8) {\r
-#if defined(STBI_SSE2)\r
- // load and perform the vertical filtering pass\r
- // this uses 3*x + y = 4*x + (y - x)\r
- __m128i zero = _mm_setzero_si128();\r
- __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i));\r
- __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i));\r
- __m128i farw = _mm_unpacklo_epi8(farb, zero);\r
- __m128i nearw = _mm_unpacklo_epi8(nearb, zero);\r
- __m128i diff = _mm_sub_epi16(farw, nearw);\r
- __m128i nears = _mm_slli_epi16(nearw, 2);\r
- __m128i curr = _mm_add_epi16(nears, diff); // current row\r
-\r
- // horizontal filter works the same based on shifted vers of current\r
- // row. "prev" is current row shifted right by 1 pixel; we need to\r
- // insert the previous pixel value (from t1).\r
- // "next" is current row shifted left by 1 pixel, with first pixel\r
- // of next block of 8 pixels added in.\r
- __m128i prv0 = _mm_slli_si128(curr, 2);\r
- __m128i nxt0 = _mm_srli_si128(curr, 2);\r
- __m128i prev = _mm_insert_epi16(prv0, t1, 0);\r
- __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7);\r
-\r
- // horizontal filter, polyphase implementation since it's convenient:\r
- // even pixels = 3*cur + prev = cur*4 + (prev - cur)\r
- // odd pixels = 3*cur + next = cur*4 + (next - cur)\r
- // note the shared term.\r
- __m128i bias = _mm_set1_epi16(8);\r
- __m128i curs = _mm_slli_epi16(curr, 2);\r
- __m128i prvd = _mm_sub_epi16(prev, curr);\r
- __m128i nxtd = _mm_sub_epi16(next, curr);\r
- __m128i curb = _mm_add_epi16(curs, bias);\r
- __m128i even = _mm_add_epi16(prvd, curb);\r
- __m128i odd = _mm_add_epi16(nxtd, curb);\r
-\r
- // interleave even and odd pixels, then undo scaling.\r
- __m128i int0 = _mm_unpacklo_epi16(even, odd);\r
- __m128i int1 = _mm_unpackhi_epi16(even, odd);\r
- __m128i de0 = _mm_srli_epi16(int0, 4);\r
- __m128i de1 = _mm_srli_epi16(int1, 4);\r
-\r
- // pack and write output\r
- __m128i outv = _mm_packus_epi16(de0, de1);\r
- _mm_storeu_si128((__m128i *) (out + i*2), outv);\r
-#elif defined(STBI_NEON)\r
- // load and perform the vertical filtering pass\r
- // this uses 3*x + y = 4*x + (y - x)\r
- uint8x8_t farb = vld1_u8(in_far + i);\r
- uint8x8_t nearb = vld1_u8(in_near + i);\r
- int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb));\r
- int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2));\r
- int16x8_t curr = vaddq_s16(nears, diff); // current row\r
-\r
- // horizontal filter works the same based on shifted vers of current\r
- // row. "prev" is current row shifted right by 1 pixel; we need to\r
- // insert the previous pixel value (from t1).\r
- // "next" is current row shifted left by 1 pixel, with first pixel\r
- // of next block of 8 pixels added in.\r
- int16x8_t prv0 = vextq_s16(curr, curr, 7);\r
- int16x8_t nxt0 = vextq_s16(curr, curr, 1);\r
- int16x8_t prev = vsetq_lane_s16(t1, prv0, 0);\r
- int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7);\r
-\r
- // horizontal filter, polyphase implementation since it's convenient:\r
- // even pixels = 3*cur + prev = cur*4 + (prev - cur)\r
- // odd pixels = 3*cur + next = cur*4 + (next - cur)\r
- // note the shared term.\r
- int16x8_t curs = vshlq_n_s16(curr, 2);\r
- int16x8_t prvd = vsubq_s16(prev, curr);\r
- int16x8_t nxtd = vsubq_s16(next, curr);\r
- int16x8_t even = vaddq_s16(curs, prvd);\r
- int16x8_t odd = vaddq_s16(curs, nxtd);\r
-\r
- // undo scaling and round, then store with even/odd phases interleaved\r
- uint8x8x2_t o;\r
- o.val[0] = vqrshrun_n_s16(even, 4);\r
- o.val[1] = vqrshrun_n_s16(odd, 4);\r
- vst2_u8(out + i*2, o);\r
-#endif\r
-\r
- // "previous" value for next iter\r
- t1 = 3*in_near[i+7] + in_far[i+7];\r
- }\r
-\r
- t0 = t1;\r
- t1 = 3*in_near[i] + in_far[i];\r
- out[i*2] = stbi__div16(3*t1 + t0 + 8);\r
-\r
- for (++i; i < w; ++i) {\r
- t0 = t1;\r
- t1 = 3*in_near[i]+in_far[i];\r
- out[i*2-1] = stbi__div16(3*t0 + t1 + 8);\r
- out[i*2 ] = stbi__div16(3*t1 + t0 + 8);\r
- }\r
- out[w*2-1] = stbi__div4(t1+2);\r
-\r
- STBI_NOTUSED(hs);\r
-\r
- return out;\r
-}\r
-#endif\r
-\r
-static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\r
-{\r
- // resample with nearest-neighbor\r
- int i,j;\r
- STBI_NOTUSED(in_far);\r
- for (i=0; i < w; ++i)\r
- for (j=0; j < hs; ++j)\r
- out[i*hs+j] = in_near[i];\r
- return out;\r
-}\r
-\r
-// this is a reduced-precision calculation of YCbCr-to-RGB introduced\r
-// to make sure the code produces the same results in both SIMD and scalar\r
-#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8)\r
-static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)\r
-{\r
- int i;\r
- for (i=0; i < count; ++i) {\r
- int y_fixed = (y[i] << 20) + (1<<19); // rounding\r
- int r,g,b;\r
- int cr = pcr[i] - 128;\r
- int cb = pcb[i] - 128;\r
- r = y_fixed + cr* stbi__float2fixed(1.40200f);\r
- g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);\r
- b = y_fixed + cb* stbi__float2fixed(1.77200f);\r
- r >>= 20;\r
- g >>= 20;\r
- b >>= 20;\r
- if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }\r
- if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }\r
- if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }\r
- out[0] = (stbi_uc)r;\r
- out[1] = (stbi_uc)g;\r
- out[2] = (stbi_uc)b;\r
- out[3] = 255;\r
- out += step;\r
- }\r
-}\r
-\r
-#if defined(STBI_SSE2) || defined(STBI_NEON)\r
-static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step)\r
-{\r
- int i = 0;\r
-\r
-#ifdef STBI_SSE2\r
- // step == 3 is pretty ugly on the final interleave, and i'm not convinced\r
- // it's useful in practice (you wouldn't use it for textures, for example).\r
- // so just accelerate step == 4 case.\r
- if (step == 4) {\r
- // this is a fairly straightforward implementation and not super-optimized.\r
- __m128i signflip = _mm_set1_epi8(-0x80);\r
- __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f));\r
- __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f));\r
- __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f));\r
- __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f));\r
- __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128);\r
- __m128i xw = _mm_set1_epi16(255); // alpha channel\r
-\r
- for (; i+7 < count; i += 8) {\r
- // load\r
- __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i));\r
- __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i));\r
- __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i));\r
- __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128\r
- __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128\r
-\r
- // unpack to short (and left-shift cr, cb by 8)\r
- __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes);\r
- __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased);\r
- __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased);\r
-\r
- // color transform\r
- __m128i yws = _mm_srli_epi16(yw, 4);\r
- __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw);\r
- __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw);\r
- __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1);\r
- __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1);\r
- __m128i rws = _mm_add_epi16(cr0, yws);\r
- __m128i gwt = _mm_add_epi16(cb0, yws);\r
- __m128i bws = _mm_add_epi16(yws, cb1);\r
- __m128i gws = _mm_add_epi16(gwt, cr1);\r
-\r
- // descale\r
- __m128i rw = _mm_srai_epi16(rws, 4);\r
- __m128i bw = _mm_srai_epi16(bws, 4);\r
- __m128i gw = _mm_srai_epi16(gws, 4);\r
-\r
- // back to byte, set up for transpose\r
- __m128i brb = _mm_packus_epi16(rw, bw);\r
- __m128i gxb = _mm_packus_epi16(gw, xw);\r
-\r
- // transpose to interleave channels\r
- __m128i t0 = _mm_unpacklo_epi8(brb, gxb);\r
- __m128i t1 = _mm_unpackhi_epi8(brb, gxb);\r
- __m128i o0 = _mm_unpacklo_epi16(t0, t1);\r
- __m128i o1 = _mm_unpackhi_epi16(t0, t1);\r
-\r
- // store\r
- _mm_storeu_si128((__m128i *) (out + 0), o0);\r
- _mm_storeu_si128((__m128i *) (out + 16), o1);\r
- out += 32;\r
- }\r
- }\r
-#endif\r
-\r
-#ifdef STBI_NEON\r
- // in this version, step=3 support would be easy to add. but is there demand?\r
- if (step == 4) {\r
- // this is a fairly straightforward implementation and not super-optimized.\r
- uint8x8_t signflip = vdup_n_u8(0x80);\r
- int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f));\r
- int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f));\r
- int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f));\r
- int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f));\r
-\r
- for (; i+7 < count; i += 8) {\r
- // load\r
- uint8x8_t y_bytes = vld1_u8(y + i);\r
- uint8x8_t cr_bytes = vld1_u8(pcr + i);\r
- uint8x8_t cb_bytes = vld1_u8(pcb + i);\r
- int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip));\r
- int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip));\r
-\r
- // expand to s16\r
- int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4));\r
- int16x8_t crw = vshll_n_s8(cr_biased, 7);\r
- int16x8_t cbw = vshll_n_s8(cb_biased, 7);\r
-\r
- // color transform\r
- int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0);\r
- int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0);\r
- int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1);\r
- int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1);\r
- int16x8_t rws = vaddq_s16(yws, cr0);\r
- int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1);\r
- int16x8_t bws = vaddq_s16(yws, cb1);\r
-\r
- // undo scaling, round, convert to byte\r
- uint8x8x4_t o;\r
- o.val[0] = vqrshrun_n_s16(rws, 4);\r
- o.val[1] = vqrshrun_n_s16(gws, 4);\r
- o.val[2] = vqrshrun_n_s16(bws, 4);\r
- o.val[3] = vdup_n_u8(255);\r
-\r
- // store, interleaving r/g/b/a\r
- vst4_u8(out, o);\r
- out += 8*4;\r
- }\r
- }\r
-#endif\r
-\r
- for (; i < count; ++i) {\r
- int y_fixed = (y[i] << 20) + (1<<19); // rounding\r
- int r,g,b;\r
- int cr = pcr[i] - 128;\r
- int cb = pcb[i] - 128;\r
- r = y_fixed + cr* stbi__float2fixed(1.40200f);\r
- g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);\r
- b = y_fixed + cb* stbi__float2fixed(1.77200f);\r
- r >>= 20;\r
- g >>= 20;\r
- b >>= 20;\r
- if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }\r
- if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }\r
- if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }\r
- out[0] = (stbi_uc)r;\r
- out[1] = (stbi_uc)g;\r
- out[2] = (stbi_uc)b;\r
- out[3] = 255;\r
- out += step;\r
- }\r
-}\r
-#endif\r
-\r
-// set up the kernels\r
-static void stbi__setup_jpeg(stbi__jpeg *j)\r
-{\r
- j->idct_block_kernel = stbi__idct_block;\r
- j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row;\r
- j->resample_row_hv_2_kernel = stbi__resample_row_hv_2;\r
-\r
-#ifdef STBI_SSE2\r
- if (stbi__sse2_available()) {\r
- j->idct_block_kernel = stbi__idct_simd;\r
- j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;\r
- j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;\r
- }\r
-#endif\r
-\r
-#ifdef STBI_NEON\r
- j->idct_block_kernel = stbi__idct_simd;\r
- j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;\r
- j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;\r
-#endif\r
-}\r
-\r
-// clean up the temporary component buffers\r
-static void stbi__cleanup_jpeg(stbi__jpeg *j)\r
-{\r
- stbi__free_jpeg_components(j, j->s->img_n, 0);\r
-}\r
-\r
-typedef struct\r
-{\r
- resample_row_func resample;\r
- stbi_uc *line0,*line1;\r
- int hs,vs; // expansion factor in each axis\r
- int w_lores; // horizontal pixels pre-expansion\r
- int ystep; // how far through vertical expansion we are\r
- int ypos; // which pre-expansion row we're on\r
-} stbi__resample;\r
-\r
-// fast 0..255 * 0..255 => 0..255 rounded multiplication\r
-static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y)\r
-{\r
- unsigned int t = x*y + 128;\r
- return (stbi_uc) ((t + (t >>8)) >> 8);\r
-}\r
-\r
-static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp)\r
-{\r
- int n, decode_n, is_rgb;\r
- z->s->img_n = 0; // make stbi__cleanup_jpeg safe\r
-\r
- // validate req_comp\r
- if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");\r
-\r
- // load a jpeg image from whichever source, but leave in YCbCr format\r
- if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; }\r
-\r
- // determine actual number of components to generate\r
- n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1;\r
-\r
- is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif));\r
-\r
- if (z->s->img_n == 3 && n < 3 && !is_rgb)\r
- decode_n = 1;\r
- else\r
- decode_n = z->s->img_n;\r
-\r
- // resample and color-convert\r
- {\r
- int k;\r
- unsigned int i,j;\r
- stbi_uc *output;\r
- stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL };\r
-\r
- stbi__resample res_comp[4];\r
-\r
- for (k=0; k < decode_n; ++k) {\r
- stbi__resample *r = &res_comp[k];\r
-\r
- // allocate line buffer big enough for upsampling off the edges\r
- // with upsample factor of 4\r
- z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3);\r
- if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }\r
-\r
- r->hs = z->img_h_max / z->img_comp[k].h;\r
- r->vs = z->img_v_max / z->img_comp[k].v;\r
- r->ystep = r->vs >> 1;\r
- r->w_lores = (z->s->img_x + r->hs-1) / r->hs;\r
- r->ypos = 0;\r
- r->line0 = r->line1 = z->img_comp[k].data;\r
-\r
- if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1;\r
- else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2;\r
- else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2;\r
- else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel;\r
- else r->resample = stbi__resample_row_generic;\r
- }\r
-\r
- // can't error after this so, this is safe\r
- output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1);\r
- if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }\r
-\r
- // now go ahead and resample\r
- for (j=0; j < z->s->img_y; ++j) {\r
- stbi_uc *out = output + n * z->s->img_x * j;\r
- for (k=0; k < decode_n; ++k) {\r
- stbi__resample *r = &res_comp[k];\r
- int y_bot = r->ystep >= (r->vs >> 1);\r
- coutput[k] = r->resample(z->img_comp[k].linebuf,\r
- y_bot ? r->line1 : r->line0,\r
- y_bot ? r->line0 : r->line1,\r
- r->w_lores, r->hs);\r
- if (++r->ystep >= r->vs) {\r
- r->ystep = 0;\r
- r->line0 = r->line1;\r
- if (++r->ypos < z->img_comp[k].y)\r
- r->line1 += z->img_comp[k].w2;\r
- }\r
- }\r
- if (n >= 3) {\r
- stbi_uc *y = coutput[0];\r
- if (z->s->img_n == 3) {\r
- if (is_rgb) {\r
- for (i=0; i < z->s->img_x; ++i) {\r
- out[0] = y[i];\r
- out[1] = coutput[1][i];\r
- out[2] = coutput[2][i];\r
- out[3] = 255;\r
- out += n;\r
- }\r
- } else {\r
- z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);\r
- }\r
- } else if (z->s->img_n == 4) {\r
- if (z->app14_color_transform == 0) { // CMYK\r
- for (i=0; i < z->s->img_x; ++i) {\r
- stbi_uc m = coutput[3][i];\r
- out[0] = stbi__blinn_8x8(coutput[0][i], m);\r
- out[1] = stbi__blinn_8x8(coutput[1][i], m);\r
- out[2] = stbi__blinn_8x8(coutput[2][i], m);\r
- out[3] = 255;\r
- out += n;\r
- }\r
- } else if (z->app14_color_transform == 2) { // YCCK\r
- z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);\r
- for (i=0; i < z->s->img_x; ++i) {\r
- stbi_uc m = coutput[3][i];\r
- out[0] = stbi__blinn_8x8(255 - out[0], m);\r
- out[1] = stbi__blinn_8x8(255 - out[1], m);\r
- out[2] = stbi__blinn_8x8(255 - out[2], m);\r
- out += n;\r
- }\r
- } else { // YCbCr + alpha? Ignore the fourth channel for now\r
- z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);\r
- }\r
- } else\r
- for (i=0; i < z->s->img_x; ++i) {\r
- out[0] = out[1] = out[2] = y[i];\r
- out[3] = 255; // not used if n==3\r
- out += n;\r
- }\r
- } else {\r
- if (is_rgb) {\r
- if (n == 1)\r
- for (i=0; i < z->s->img_x; ++i)\r
- *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);\r
- else {\r
- for (i=0; i < z->s->img_x; ++i, out += 2) {\r
- out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);\r
- out[1] = 255;\r
- }\r
- }\r
- } else if (z->s->img_n == 4 && z->app14_color_transform == 0) {\r
- for (i=0; i < z->s->img_x; ++i) {\r
- stbi_uc m = coutput[3][i];\r
- stbi_uc r = stbi__blinn_8x8(coutput[0][i], m);\r
- stbi_uc g = stbi__blinn_8x8(coutput[1][i], m);\r
- stbi_uc b = stbi__blinn_8x8(coutput[2][i], m);\r
- out[0] = stbi__compute_y(r, g, b);\r
- out[1] = 255;\r
- out += n;\r
- }\r
- } else if (z->s->img_n == 4 && z->app14_color_transform == 2) {\r
- for (i=0; i < z->s->img_x; ++i) {\r
- out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]);\r
- out[1] = 255;\r
- out += n;\r
- }\r
- } else {\r
- stbi_uc *y = coutput[0];\r
- if (n == 1)\r
- for (i=0; i < z->s->img_x; ++i) out[i] = y[i];\r
- else\r
- for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; }\r
- }\r
- }\r
- }\r
- stbi__cleanup_jpeg(z);\r
- *out_x = z->s->img_x;\r
- *out_y = z->s->img_y;\r
- if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output\r
- return output;\r
- }\r
-}\r
-\r
-static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- unsigned char* result;\r
- stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg));\r
- STBI_NOTUSED(ri);\r
- j->s = s;\r
- stbi__setup_jpeg(j);\r
- result = load_jpeg_image(j, x,y,comp,req_comp);\r
- STBI_FREE(j);\r
- return result;\r
-}\r
-\r
-static int stbi__jpeg_test(stbi__context *s)\r
-{\r
- int r;\r
- stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg));\r
- j->s = s;\r
- stbi__setup_jpeg(j);\r
- r = stbi__decode_jpeg_header(j, STBI__SCAN_type);\r
- stbi__rewind(s);\r
- STBI_FREE(j);\r
- return r;\r
-}\r
-\r
-static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp)\r
-{\r
- if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) {\r
- stbi__rewind( j->s );\r
- return 0;\r
- }\r
- if (x) *x = j->s->img_x;\r
- if (y) *y = j->s->img_y;\r
- if (comp) *comp = j->s->img_n >= 3 ? 3 : 1;\r
- return 1;\r
-}\r
-\r
-static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- int result;\r
- stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg)));\r
- j->s = s;\r
- result = stbi__jpeg_info_raw(j, x, y, comp);\r
- STBI_FREE(j);\r
- return result;\r
-}\r
-#endif\r
-\r
-// public domain zlib decode v0.2 Sean Barrett 2006-11-18\r
-// simple implementation\r
-// - all input must be provided in an upfront buffer\r
-// - all output is written to a single output buffer (can malloc/realloc)\r
-// performance\r
-// - fast huffman\r
-\r
-#ifndef STBI_NO_ZLIB\r
-\r
-// fast-way is faster to check than jpeg huffman, but slow way is slower\r
-#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables\r
-#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1)\r
-\r
-// zlib-style huffman encoding\r
-// (jpegs packs from left, zlib from right, so can't share code)\r
-typedef struct\r
-{\r
- stbi__uint16 fast[1 << STBI__ZFAST_BITS];\r
- stbi__uint16 firstcode[16];\r
- int maxcode[17];\r
- stbi__uint16 firstsymbol[16];\r
- stbi_uc size[288];\r
- stbi__uint16 value[288];\r
-} stbi__zhuffman;\r
-\r
-stbi_inline static int stbi__bitreverse16(int n)\r
-{\r
- n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1);\r
- n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2);\r
- n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4);\r
- n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8);\r
- return n;\r
-}\r
-\r
-stbi_inline static int stbi__bit_reverse(int v, int bits)\r
-{\r
- STBI_ASSERT(bits <= 16);\r
- // to bit reverse n bits, reverse 16 and shift\r
- // e.g. 11 bits, bit reverse and shift away 5\r
- return stbi__bitreverse16(v) >> (16-bits);\r
-}\r
-\r
-static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num)\r
-{\r
- int i,k=0;\r
- int code, next_code[16], sizes[17];\r
-\r
- // DEFLATE spec for generating codes\r
- memset(sizes, 0, sizeof(sizes));\r
- memset(z->fast, 0, sizeof(z->fast));\r
- for (i=0; i < num; ++i)\r
- ++sizes[sizelist[i]];\r
- sizes[0] = 0;\r
- for (i=1; i < 16; ++i)\r
- if (sizes[i] > (1 << i))\r
- return stbi__err("bad sizes", "Corrupt PNG");\r
- code = 0;\r
- for (i=1; i < 16; ++i) {\r
- next_code[i] = code;\r
- z->firstcode[i] = (stbi__uint16) code;\r
- z->firstsymbol[i] = (stbi__uint16) k;\r
- code = (code + sizes[i]);\r
- if (sizes[i])\r
- if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG");\r
- z->maxcode[i] = code << (16-i); // preshift for inner loop\r
- code <<= 1;\r
- k += sizes[i];\r
- }\r
- z->maxcode[16] = 0x10000; // sentinel\r
- for (i=0; i < num; ++i) {\r
- int s = sizelist[i];\r
- if (s) {\r
- int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];\r
- stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i);\r
- z->size [c] = (stbi_uc ) s;\r
- z->value[c] = (stbi__uint16) i;\r
- if (s <= STBI__ZFAST_BITS) {\r
- int j = stbi__bit_reverse(next_code[s],s);\r
- while (j < (1 << STBI__ZFAST_BITS)) {\r
- z->fast[j] = fastv;\r
- j += (1 << s);\r
- }\r
- }\r
- ++next_code[s];\r
- }\r
- }\r
- return 1;\r
-}\r
-\r
-// zlib-from-memory implementation for PNG reading\r
-// because PNG allows splitting the zlib stream arbitrarily,\r
-// and it's annoying structurally to have PNG call ZLIB call PNG,\r
-// we require PNG read all the IDATs and combine them into a single\r
-// memory buffer\r
-\r
-typedef struct\r
-{\r
- stbi_uc *zbuffer, *zbuffer_end;\r
- int num_bits;\r
- stbi__uint32 code_buffer;\r
-\r
- char *zout;\r
- char *zout_start;\r
- char *zout_end;\r
- int z_expandable;\r
-\r
- stbi__zhuffman z_length, z_distance;\r
-} stbi__zbuf;\r
-\r
-stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z)\r
-{\r
- if (z->zbuffer >= z->zbuffer_end) return 0;\r
- return *z->zbuffer++;\r
-}\r
-\r
-static void stbi__fill_bits(stbi__zbuf *z)\r
-{\r
- do {\r
- STBI_ASSERT(z->code_buffer < (1U << z->num_bits));\r
- z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits;\r
- z->num_bits += 8;\r
- } while (z->num_bits <= 24);\r
-}\r
-\r
-stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n)\r
-{\r
- unsigned int k;\r
- if (z->num_bits < n) stbi__fill_bits(z);\r
- k = z->code_buffer & ((1 << n) - 1);\r
- z->code_buffer >>= n;\r
- z->num_bits -= n;\r
- return k;\r
-}\r
-\r
-static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)\r
-{\r
- int b,s,k;\r
- // not resolved by fast table, so compute it the slow way\r
- // use jpeg approach, which requires MSbits at top\r
- k = stbi__bit_reverse(a->code_buffer, 16);\r
- for (s=STBI__ZFAST_BITS+1; ; ++s)\r
- if (k < z->maxcode[s])\r
- break;\r
- if (s == 16) return -1; // invalid code!\r
- // code size is s, so:\r
- b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];\r
- STBI_ASSERT(z->size[b] == s);\r
- a->code_buffer >>= s;\r
- a->num_bits -= s;\r
- return z->value[b];\r
-}\r
-\r
-stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)\r
-{\r
- int b,s;\r
- if (a->num_bits < 16) stbi__fill_bits(a);\r
- b = z->fast[a->code_buffer & STBI__ZFAST_MASK];\r
- if (b) {\r
- s = b >> 9;\r
- a->code_buffer >>= s;\r
- a->num_bits -= s;\r
- return b & 511;\r
- }\r
- return stbi__zhuffman_decode_slowpath(a, z);\r
-}\r
-\r
-static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes\r
-{\r
- char *q;\r
- int cur, limit, old_limit;\r
- z->zout = zout;\r
- if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG");\r
- cur = (int) (z->zout - z->zout_start);\r
- limit = old_limit = (int) (z->zout_end - z->zout_start);\r
- while (cur + n > limit)\r
- limit *= 2;\r
- q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit);\r
- STBI_NOTUSED(old_limit);\r
- if (q == NULL) return stbi__err("outofmem", "Out of memory");\r
- z->zout_start = q;\r
- z->zout = q + cur;\r
- z->zout_end = q + limit;\r
- return 1;\r
-}\r
-\r
-static const int stbi__zlength_base[31] = {\r
- 3,4,5,6,7,8,9,10,11,13,\r
- 15,17,19,23,27,31,35,43,51,59,\r
- 67,83,99,115,131,163,195,227,258,0,0 };\r
-\r
-static const int stbi__zlength_extra[31]=\r
-{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };\r
-\r
-static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,\r
-257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};\r
-\r
-static const int stbi__zdist_extra[32] =\r
-{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};\r
-\r
-static int stbi__parse_huffman_block(stbi__zbuf *a)\r
-{\r
- char *zout = a->zout;\r
- for(;;) {\r
- int z = stbi__zhuffman_decode(a, &a->z_length);\r
- if (z < 256) {\r
- if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes\r
- if (zout >= a->zout_end) {\r
- if (!stbi__zexpand(a, zout, 1)) return 0;\r
- zout = a->zout;\r
- }\r
- *zout++ = (char) z;\r
- } else {\r
- stbi_uc *p;\r
- int len,dist;\r
- if (z == 256) {\r
- a->zout = zout;\r
- return 1;\r
- }\r
- z -= 257;\r
- len = stbi__zlength_base[z];\r
- if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);\r
- z = stbi__zhuffman_decode(a, &a->z_distance);\r
- if (z < 0) return stbi__err("bad huffman code","Corrupt PNG");\r
- dist = stbi__zdist_base[z];\r
- if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);\r
- if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG");\r
- if (zout + len > a->zout_end) {\r
- if (!stbi__zexpand(a, zout, len)) return 0;\r
- zout = a->zout;\r
- }\r
- p = (stbi_uc *) (zout - dist);\r
- if (dist == 1) { // run of one byte; common in images.\r
- stbi_uc v = *p;\r
- if (len) { do *zout++ = v; while (--len); }\r
- } else {\r
- if (len) { do *zout++ = *p++; while (--len); }\r
- }\r
- }\r
- }\r
-}\r
-\r
-static int stbi__compute_huffman_codes(stbi__zbuf *a)\r
-{\r
- static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };\r
- stbi__zhuffman z_codelength;\r
- stbi_uc lencodes[286+32+137];//padding for maximum single op\r
- stbi_uc codelength_sizes[19];\r
- int i,n;\r
-\r
- int hlit = stbi__zreceive(a,5) + 257;\r
- int hdist = stbi__zreceive(a,5) + 1;\r
- int hclen = stbi__zreceive(a,4) + 4;\r
- int ntot = hlit + hdist;\r
-\r
- memset(codelength_sizes, 0, sizeof(codelength_sizes));\r
- for (i=0; i < hclen; ++i) {\r
- int s = stbi__zreceive(a,3);\r
- codelength_sizes[length_dezigzag[i]] = (stbi_uc) s;\r
- }\r
- if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;\r
-\r
- n = 0;\r
- while (n < ntot) {\r
- int c = stbi__zhuffman_decode(a, &z_codelength);\r
- if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG");\r
- if (c < 16)\r
- lencodes[n++] = (stbi_uc) c;\r
- else {\r
- stbi_uc fill = 0;\r
- if (c == 16) {\r
- c = stbi__zreceive(a,2)+3;\r
- if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG");\r
- fill = lencodes[n-1];\r
- } else if (c == 17)\r
- c = stbi__zreceive(a,3)+3;\r
- else {\r
- STBI_ASSERT(c == 18);\r
- c = stbi__zreceive(a,7)+11;\r
- }\r
- if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG");\r
- memset(lencodes+n, fill, c);\r
- n += c;\r
- }\r
- }\r
- if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG");\r
- if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;\r
- if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;\r
- return 1;\r
-}\r
-\r
-static int stbi__parse_uncompressed_block(stbi__zbuf *a)\r
-{\r
- stbi_uc header[4];\r
- int len,nlen,k;\r
- if (a->num_bits & 7)\r
- stbi__zreceive(a, a->num_bits & 7); // discard\r
- // drain the bit-packed data into header\r
- k = 0;\r
- while (a->num_bits > 0) {\r
- header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check\r
- a->code_buffer >>= 8;\r
- a->num_bits -= 8;\r
- }\r
- STBI_ASSERT(a->num_bits == 0);\r
- // now fill header the normal way\r
- while (k < 4)\r
- header[k++] = stbi__zget8(a);\r
- len = header[1] * 256 + header[0];\r
- nlen = header[3] * 256 + header[2];\r
- if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG");\r
- if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG");\r
- if (a->zout + len > a->zout_end)\r
- if (!stbi__zexpand(a, a->zout, len)) return 0;\r
- memcpy(a->zout, a->zbuffer, len);\r
- a->zbuffer += len;\r
- a->zout += len;\r
- return 1;\r
-}\r
-\r
-static int stbi__parse_zlib_header(stbi__zbuf *a)\r
-{\r
- int cmf = stbi__zget8(a);\r
- int cm = cmf & 15;\r
- /* int cinfo = cmf >> 4; */\r
- int flg = stbi__zget8(a);\r
- if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec\r
- if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png\r
- if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png\r
- // window = 1 << (8 + cinfo)... but who cares, we fully buffer output\r
- return 1;\r
-}\r
-\r
-static const stbi_uc stbi__zdefault_length[288] =\r
-{\r
- 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\r
- 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\r
- 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\r
- 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\r
- 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,\r
- 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,\r
- 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,\r
- 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,\r
- 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8\r
-};\r
-static const stbi_uc stbi__zdefault_distance[32] =\r
-{\r
- 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5\r
-};\r
-/*\r
-Init algorithm:\r
-{\r
- int i; // use <= to match clearly with spec\r
- for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8;\r
- for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9;\r
- for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7;\r
- for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8;\r
-\r
- for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5;\r
-}\r
-*/\r
-\r
-static int stbi__parse_zlib(stbi__zbuf *a, int parse_header)\r
-{\r
- int final, type;\r
- if (parse_header)\r
- if (!stbi__parse_zlib_header(a)) return 0;\r
- a->num_bits = 0;\r
- a->code_buffer = 0;\r
- do {\r
- final = stbi__zreceive(a,1);\r
- type = stbi__zreceive(a,2);\r
- if (type == 0) {\r
- if (!stbi__parse_uncompressed_block(a)) return 0;\r
- } else if (type == 3) {\r
- return 0;\r
- } else {\r
- if (type == 1) {\r
- // use fixed code lengths\r
- if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , 288)) return 0;\r
- if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0;\r
- } else {\r
- if (!stbi__compute_huffman_codes(a)) return 0;\r
- }\r
- if (!stbi__parse_huffman_block(a)) return 0;\r
- }\r
- } while (!final);\r
- return 1;\r
-}\r
-\r
-static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)\r
-{\r
- a->zout_start = obuf;\r
- a->zout = obuf;\r
- a->zout_end = obuf + olen;\r
- a->z_expandable = exp;\r
-\r
- return stbi__parse_zlib(a, parse_header);\r
-}\r
-\r
-STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)\r
-{\r
- stbi__zbuf a;\r
- char *p = (char *) stbi__malloc(initial_size);\r
- if (p == NULL) return NULL;\r
- a.zbuffer = (stbi_uc *) buffer;\r
- a.zbuffer_end = (stbi_uc *) buffer + len;\r
- if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {\r
- if (outlen) *outlen = (int) (a.zout - a.zout_start);\r
- return a.zout_start;\r
- } else {\r
- STBI_FREE(a.zout_start);\r
- return NULL;\r
- }\r
-}\r
-\r
-STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)\r
-{\r
- return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);\r
-}\r
-\r
-STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)\r
-{\r
- stbi__zbuf a;\r
- char *p = (char *) stbi__malloc(initial_size);\r
- if (p == NULL) return NULL;\r
- a.zbuffer = (stbi_uc *) buffer;\r
- a.zbuffer_end = (stbi_uc *) buffer + len;\r
- if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {\r
- if (outlen) *outlen = (int) (a.zout - a.zout_start);\r
- return a.zout_start;\r
- } else {\r
- STBI_FREE(a.zout_start);\r
- return NULL;\r
- }\r
-}\r
-\r
-STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)\r
-{\r
- stbi__zbuf a;\r
- a.zbuffer = (stbi_uc *) ibuffer;\r
- a.zbuffer_end = (stbi_uc *) ibuffer + ilen;\r
- if (stbi__do_zlib(&a, obuffer, olen, 0, 1))\r
- return (int) (a.zout - a.zout_start);\r
- else\r
- return -1;\r
-}\r
-\r
-STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)\r
-{\r
- stbi__zbuf a;\r
- char *p = (char *) stbi__malloc(16384);\r
- if (p == NULL) return NULL;\r
- a.zbuffer = (stbi_uc *) buffer;\r
- a.zbuffer_end = (stbi_uc *) buffer+len;\r
- if (stbi__do_zlib(&a, p, 16384, 1, 0)) {\r
- if (outlen) *outlen = (int) (a.zout - a.zout_start);\r
- return a.zout_start;\r
- } else {\r
- STBI_FREE(a.zout_start);\r
- return NULL;\r
- }\r
-}\r
-\r
-STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)\r
-{\r
- stbi__zbuf a;\r
- a.zbuffer = (stbi_uc *) ibuffer;\r
- a.zbuffer_end = (stbi_uc *) ibuffer + ilen;\r
- if (stbi__do_zlib(&a, obuffer, olen, 0, 0))\r
- return (int) (a.zout - a.zout_start);\r
- else\r
- return -1;\r
-}\r
-#endif\r
-\r
-// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18\r
-// simple implementation\r
-// - only 8-bit samples\r
-// - no CRC checking\r
-// - allocates lots of intermediate memory\r
-// - avoids problem of streaming data between subsystems\r
-// - avoids explicit window management\r
-// performance\r
-// - uses stb_zlib, a PD zlib implementation with fast huffman decoding\r
-\r
-#ifndef STBI_NO_PNG\r
-typedef struct\r
-{\r
- stbi__uint32 length;\r
- stbi__uint32 type;\r
-} stbi__pngchunk;\r
-\r
-static stbi__pngchunk stbi__get_chunk_header(stbi__context *s)\r
-{\r
- stbi__pngchunk c;\r
- c.length = stbi__get32be(s);\r
- c.type = stbi__get32be(s);\r
- return c;\r
-}\r
-\r
-static int stbi__check_png_header(stbi__context *s)\r
-{\r
- static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 };\r
- int i;\r
- for (i=0; i < 8; ++i)\r
- if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG");\r
- return 1;\r
-}\r
-\r
-typedef struct\r
-{\r
- stbi__context *s;\r
- stbi_uc *idata, *expanded, *out;\r
- int depth;\r
-} stbi__png;\r
-\r
-\r
-enum {\r
- STBI__F_none=0,\r
- STBI__F_sub=1,\r
- STBI__F_up=2,\r
- STBI__F_avg=3,\r
- STBI__F_paeth=4,\r
- // synthetic filters used for first scanline to avoid needing a dummy row of 0s\r
- STBI__F_avg_first,\r
- STBI__F_paeth_first\r
-};\r
-\r
-static stbi_uc first_row_filter[5] =\r
-{\r
- STBI__F_none,\r
- STBI__F_sub,\r
- STBI__F_none,\r
- STBI__F_avg_first,\r
- STBI__F_paeth_first\r
-};\r
-\r
-static int stbi__paeth(int a, int b, int c)\r
-{\r
- int p = a + b - c;\r
- int pa = abs(p-a);\r
- int pb = abs(p-b);\r
- int pc = abs(p-c);\r
- if (pa <= pb && pa <= pc) return a;\r
- if (pb <= pc) return b;\r
- return c;\r
-}\r
-\r
-static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 };\r
-\r
-// create the png data from post-deflated data\r
-static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color)\r
-{\r
- int bytes = (depth == 16? 2 : 1);\r
- stbi__context *s = a->s;\r
- stbi__uint32 i,j,stride = x*out_n*bytes;\r
- stbi__uint32 img_len, img_width_bytes;\r
- int k;\r
- int img_n = s->img_n; // copy it into a local for later\r
-\r
- int output_bytes = out_n*bytes;\r
- int filter_bytes = img_n*bytes;\r
- int width = x;\r
-\r
- STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1);\r
- a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into\r
- if (!a->out) return stbi__err("outofmem", "Out of memory");\r
-\r
- if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG");\r
- img_width_bytes = (((img_n * x * depth) + 7) >> 3);\r
- img_len = (img_width_bytes + 1) * y;\r
-\r
- // we used to check for exact match between raw_len and img_len on non-interlaced PNGs,\r
- // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),\r
- // so just check for raw_len < img_len always.\r
- if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");\r
-\r
- for (j=0; j < y; ++j) {\r
- stbi_uc *cur = a->out + stride*j;\r
- stbi_uc *prior;\r
- int filter = *raw++;\r
-\r
- if (filter > 4)\r
- return stbi__err("invalid filter","Corrupt PNG");\r
-\r
- if (depth < 8) {\r
- STBI_ASSERT(img_width_bytes <= x);\r
- cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place\r
- filter_bytes = 1;\r
- width = img_width_bytes;\r
- }\r
- prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above\r
-\r
- // if first row, use special filter that doesn't sample previous row\r
- if (j == 0) filter = first_row_filter[filter];\r
-\r
- // handle first byte explicitly\r
- for (k=0; k < filter_bytes; ++k) {\r
- switch (filter) {\r
- case STBI__F_none : cur[k] = raw[k]; break;\r
- case STBI__F_sub : cur[k] = raw[k]; break;\r
- case STBI__F_up : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;\r
- case STBI__F_avg : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break;\r
- case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;\r
- case STBI__F_avg_first : cur[k] = raw[k]; break;\r
- case STBI__F_paeth_first: cur[k] = raw[k]; break;\r
- }\r
- }\r
-\r
- if (depth == 8) {\r
- if (img_n != out_n)\r
- cur[img_n] = 255; // first pixel\r
- raw += img_n;\r
- cur += out_n;\r
- prior += out_n;\r
- } else if (depth == 16) {\r
- if (img_n != out_n) {\r
- cur[filter_bytes] = 255; // first pixel top byte\r
- cur[filter_bytes+1] = 255; // first pixel bottom byte\r
- }\r
- raw += filter_bytes;\r
- cur += output_bytes;\r
- prior += output_bytes;\r
- } else {\r
- raw += 1;\r
- cur += 1;\r
- prior += 1;\r
- }\r
-\r
- // this is a little gross, so that we don't switch per-pixel or per-component\r
- if (depth < 8 || img_n == out_n) {\r
- int nk = (width - 1)*filter_bytes;\r
- #define STBI__CASE(f) \\r
- case f: \\r
- for (k=0; k < nk; ++k)\r
- switch (filter) {\r
- // "none" filter turns into a memcpy here; make that explicit.\r
- case STBI__F_none: memcpy(cur, raw, nk); break;\r
- STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break;\r
- STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;\r
- STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break;\r
- STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break;\r
- STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break;\r
- STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break;\r
- }\r
- #undef STBI__CASE\r
- raw += nk;\r
- } else {\r
- STBI_ASSERT(img_n+1 == out_n);\r
- #define STBI__CASE(f) \\r
- case f: \\r
- for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \\r
- for (k=0; k < filter_bytes; ++k)\r
- switch (filter) {\r
- STBI__CASE(STBI__F_none) { cur[k] = raw[k]; } break;\r
- STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break;\r
- STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;\r
- STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break;\r
- STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break;\r
- STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break;\r
- STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break;\r
- }\r
- #undef STBI__CASE\r
-\r
- // the loop above sets the high byte of the pixels' alpha, but for\r
- // 16 bit png files we also need the low byte set. we'll do that here.\r
- if (depth == 16) {\r
- cur = a->out + stride*j; // start at the beginning of the row again\r
- for (i=0; i < x; ++i,cur+=output_bytes) {\r
- cur[filter_bytes+1] = 255;\r
- }\r
- }\r
- }\r
- }\r
-\r
- // we make a separate pass to expand bits to pixels; for performance,\r
- // this could run two scanlines behind the above code, so it won't\r
- // intefere with filtering but will still be in the cache.\r
- if (depth < 8) {\r
- for (j=0; j < y; ++j) {\r
- stbi_uc *cur = a->out + stride*j;\r
- stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes;\r
- // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit\r
- // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop\r
- stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range\r
-\r
- // note that the final byte might overshoot and write more data than desired.\r
- // we can allocate enough data that this never writes out of memory, but it\r
- // could also overwrite the next scanline. can it overwrite non-empty data\r
- // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel.\r
- // so we need to explicitly clamp the final ones\r
-\r
- if (depth == 4) {\r
- for (k=x*img_n; k >= 2; k-=2, ++in) {\r
- *cur++ = scale * ((*in >> 4) );\r
- *cur++ = scale * ((*in ) & 0x0f);\r
- }\r
- if (k > 0) *cur++ = scale * ((*in >> 4) );\r
- } else if (depth == 2) {\r
- for (k=x*img_n; k >= 4; k-=4, ++in) {\r
- *cur++ = scale * ((*in >> 6) );\r
- *cur++ = scale * ((*in >> 4) & 0x03);\r
- *cur++ = scale * ((*in >> 2) & 0x03);\r
- *cur++ = scale * ((*in ) & 0x03);\r
- }\r
- if (k > 0) *cur++ = scale * ((*in >> 6) );\r
- if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03);\r
- if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03);\r
- } else if (depth == 1) {\r
- for (k=x*img_n; k >= 8; k-=8, ++in) {\r
- *cur++ = scale * ((*in >> 7) );\r
- *cur++ = scale * ((*in >> 6) & 0x01);\r
- *cur++ = scale * ((*in >> 5) & 0x01);\r
- *cur++ = scale * ((*in >> 4) & 0x01);\r
- *cur++ = scale * ((*in >> 3) & 0x01);\r
- *cur++ = scale * ((*in >> 2) & 0x01);\r
- *cur++ = scale * ((*in >> 1) & 0x01);\r
- *cur++ = scale * ((*in ) & 0x01);\r
- }\r
- if (k > 0) *cur++ = scale * ((*in >> 7) );\r
- if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01);\r
- if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01);\r
- if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01);\r
- if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01);\r
- if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01);\r
- if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01);\r
- }\r
- if (img_n != out_n) {\r
- int q;\r
- // insert alpha = 255\r
- cur = a->out + stride*j;\r
- if (img_n == 1) {\r
- for (q=x-1; q >= 0; --q) {\r
- cur[q*2+1] = 255;\r
- cur[q*2+0] = cur[q];\r
- }\r
- } else {\r
- STBI_ASSERT(img_n == 3);\r
- for (q=x-1; q >= 0; --q) {\r
- cur[q*4+3] = 255;\r
- cur[q*4+2] = cur[q*3+2];\r
- cur[q*4+1] = cur[q*3+1];\r
- cur[q*4+0] = cur[q*3+0];\r
- }\r
- }\r
- }\r
- }\r
- } else if (depth == 16) {\r
- // force the image data from big-endian to platform-native.\r
- // this is done in a separate pass due to the decoding relying\r
- // on the data being untouched, but could probably be done\r
- // per-line during decode if care is taken.\r
- stbi_uc *cur = a->out;\r
- stbi__uint16 *cur16 = (stbi__uint16*)cur;\r
-\r
- for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) {\r
- *cur16 = (cur[0] << 8) | cur[1];\r
- }\r
- }\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced)\r
-{\r
- int bytes = (depth == 16 ? 2 : 1);\r
- int out_bytes = out_n * bytes;\r
- stbi_uc *final;\r
- int p;\r
- if (!interlaced)\r
- return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);\r
-\r
- // de-interlacing\r
- final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0);\r
- for (p=0; p < 7; ++p) {\r
- int xorig[] = { 0,4,0,2,0,1,0 };\r
- int yorig[] = { 0,0,4,0,2,0,1 };\r
- int xspc[] = { 8,8,4,4,2,2,1 };\r
- int yspc[] = { 8,8,8,4,4,2,2 };\r
- int i,j,x,y;\r
- // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1\r
- x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];\r
- y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];\r
- if (x && y) {\r
- stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;\r
- if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {\r
- STBI_FREE(final);\r
- return 0;\r
- }\r
- for (j=0; j < y; ++j) {\r
- for (i=0; i < x; ++i) {\r
- int out_y = j*yspc[p]+yorig[p];\r
- int out_x = i*xspc[p]+xorig[p];\r
- memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes,\r
- a->out + (j*x+i)*out_bytes, out_bytes);\r
- }\r
- }\r
- STBI_FREE(a->out);\r
- image_data += img_len;\r
- image_data_len -= img_len;\r
- }\r
- }\r
- a->out = final;\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n)\r
-{\r
- stbi__context *s = z->s;\r
- stbi__uint32 i, pixel_count = s->img_x * s->img_y;\r
- stbi_uc *p = z->out;\r
-\r
- // compute color-based transparency, assuming we've\r
- // already got 255 as the alpha value in the output\r
- STBI_ASSERT(out_n == 2 || out_n == 4);\r
-\r
- if (out_n == 2) {\r
- for (i=0; i < pixel_count; ++i) {\r
- p[1] = (p[0] == tc[0] ? 0 : 255);\r
- p += 2;\r
- }\r
- } else {\r
- for (i=0; i < pixel_count; ++i) {\r
- if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])\r
- p[3] = 0;\r
- p += 4;\r
- }\r
- }\r
- return 1;\r
-}\r
-\r
-static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n)\r
-{\r
- stbi__context *s = z->s;\r
- stbi__uint32 i, pixel_count = s->img_x * s->img_y;\r
- stbi__uint16 *p = (stbi__uint16*) z->out;\r
-\r
- // compute color-based transparency, assuming we've\r
- // already got 65535 as the alpha value in the output\r
- STBI_ASSERT(out_n == 2 || out_n == 4);\r
-\r
- if (out_n == 2) {\r
- for (i = 0; i < pixel_count; ++i) {\r
- p[1] = (p[0] == tc[0] ? 0 : 65535);\r
- p += 2;\r
- }\r
- } else {\r
- for (i = 0; i < pixel_count; ++i) {\r
- if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])\r
- p[3] = 0;\r
- p += 4;\r
- }\r
- }\r
- return 1;\r
-}\r
-\r
-static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n)\r
-{\r
- stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y;\r
- stbi_uc *p, *temp_out, *orig = a->out;\r
-\r
- p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0);\r
- if (p == NULL) return stbi__err("outofmem", "Out of memory");\r
-\r
- // between here and free(out) below, exitting would leak\r
- temp_out = p;\r
-\r
- if (pal_img_n == 3) {\r
- for (i=0; i < pixel_count; ++i) {\r
- int n = orig[i]*4;\r
- p[0] = palette[n ];\r
- p[1] = palette[n+1];\r
- p[2] = palette[n+2];\r
- p += 3;\r
- }\r
- } else {\r
- for (i=0; i < pixel_count; ++i) {\r
- int n = orig[i]*4;\r
- p[0] = palette[n ];\r
- p[1] = palette[n+1];\r
- p[2] = palette[n+2];\r
- p[3] = palette[n+3];\r
- p += 4;\r
- }\r
- }\r
- STBI_FREE(a->out);\r
- a->out = temp_out;\r
-\r
- STBI_NOTUSED(len);\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__unpremultiply_on_load = 0;\r
-static int stbi__de_iphone_flag = 0;\r
-\r
-STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply)\r
-{\r
- stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply;\r
-}\r
-\r
-STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert)\r
-{\r
- stbi__de_iphone_flag = flag_true_if_should_convert;\r
-}\r
-\r
-static void stbi__de_iphone(stbi__png *z)\r
-{\r
- stbi__context *s = z->s;\r
- stbi__uint32 i, pixel_count = s->img_x * s->img_y;\r
- stbi_uc *p = z->out;\r
-\r
- if (s->img_out_n == 3) { // convert bgr to rgb\r
- for (i=0; i < pixel_count; ++i) {\r
- stbi_uc t = p[0];\r
- p[0] = p[2];\r
- p[2] = t;\r
- p += 3;\r
- }\r
- } else {\r
- STBI_ASSERT(s->img_out_n == 4);\r
- if (stbi__unpremultiply_on_load) {\r
- // convert bgr to rgb and unpremultiply\r
- for (i=0; i < pixel_count; ++i) {\r
- stbi_uc a = p[3];\r
- stbi_uc t = p[0];\r
- if (a) {\r
- stbi_uc half = a / 2;\r
- p[0] = (p[2] * 255 + half) / a;\r
- p[1] = (p[1] * 255 + half) / a;\r
- p[2] = ( t * 255 + half) / a;\r
- } else {\r
- p[0] = p[2];\r
- p[2] = t;\r
- }\r
- p += 4;\r
- }\r
- } else {\r
- // convert bgr to rgb\r
- for (i=0; i < pixel_count; ++i) {\r
- stbi_uc t = p[0];\r
- p[0] = p[2];\r
- p[2] = t;\r
- p += 4;\r
- }\r
- }\r
- }\r
-}\r
-\r
-#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d))\r
-\r
-static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)\r
-{\r
- stbi_uc palette[1024], pal_img_n=0;\r
- stbi_uc has_trans=0, tc[3]={0};\r
- stbi__uint16 tc16[3];\r
- stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0;\r
- int first=1,k,interlace=0, color=0, is_iphone=0;\r
- stbi__context *s = z->s;\r
-\r
- z->expanded = NULL;\r
- z->idata = NULL;\r
- z->out = NULL;\r
-\r
- if (!stbi__check_png_header(s)) return 0;\r
-\r
- if (scan == STBI__SCAN_type) return 1;\r
-\r
- for (;;) {\r
- stbi__pngchunk c = stbi__get_chunk_header(s);\r
- switch (c.type) {\r
- case STBI__PNG_TYPE('C','g','B','I'):\r
- is_iphone = 1;\r
- stbi__skip(s, c.length);\r
- break;\r
- case STBI__PNG_TYPE('I','H','D','R'): {\r
- int comp,filter;\r
- if (!first) return stbi__err("multiple IHDR","Corrupt PNG");\r
- first = 0;\r
- if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG");\r
- s->img_x = stbi__get32be(s); if (s->img_x > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");\r
- s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");\r
- z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only");\r
- color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG");\r
- if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG");\r
- if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG");\r
- comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG");\r
- filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG");\r
- interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG");\r
- if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG");\r
- if (!pal_img_n) {\r
- s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);\r
- if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode");\r
- if (scan == STBI__SCAN_header) return 1;\r
- } else {\r
- // if paletted, then pal_n is our final components, and\r
- // img_n is # components to decompress/filter.\r
- s->img_n = 1;\r
- if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG");\r
- // if SCAN_header, have to scan to see if we have a tRNS\r
- }\r
- break;\r
- }\r
-\r
- case STBI__PNG_TYPE('P','L','T','E'): {\r
- if (first) return stbi__err("first not IHDR", "Corrupt PNG");\r
- if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG");\r
- pal_len = c.length / 3;\r
- if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG");\r
- for (i=0; i < pal_len; ++i) {\r
- palette[i*4+0] = stbi__get8(s);\r
- palette[i*4+1] = stbi__get8(s);\r
- palette[i*4+2] = stbi__get8(s);\r
- palette[i*4+3] = 255;\r
- }\r
- break;\r
- }\r
-\r
- case STBI__PNG_TYPE('t','R','N','S'): {\r
- if (first) return stbi__err("first not IHDR", "Corrupt PNG");\r
- if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG");\r
- if (pal_img_n) {\r
- if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }\r
- if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG");\r
- if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG");\r
- pal_img_n = 4;\r
- for (i=0; i < c.length; ++i)\r
- palette[i*4+3] = stbi__get8(s);\r
- } else {\r
- if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG");\r
- if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG");\r
- has_trans = 1;\r
- if (z->depth == 16) {\r
- for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is\r
- } else {\r
- for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger\r
- }\r
- }\r
- break;\r
- }\r
-\r
- case STBI__PNG_TYPE('I','D','A','T'): {\r
- if (first) return stbi__err("first not IHDR", "Corrupt PNG");\r
- if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG");\r
- if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; }\r
- if ((int)(ioff + c.length) < (int)ioff) return 0;\r
- if (ioff + c.length > idata_limit) {\r
- stbi__uint32 idata_limit_old = idata_limit;\r
- stbi_uc *p;\r
- if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;\r
- while (ioff + c.length > idata_limit)\r
- idata_limit *= 2;\r
- STBI_NOTUSED(idata_limit_old);\r
- p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory");\r
- z->idata = p;\r
- }\r
- if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG");\r
- ioff += c.length;\r
- break;\r
- }\r
-\r
- case STBI__PNG_TYPE('I','E','N','D'): {\r
- stbi__uint32 raw_len, bpl;\r
- if (first) return stbi__err("first not IHDR", "Corrupt PNG");\r
- if (scan != STBI__SCAN_load) return 1;\r
- if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG");\r
- // initial guess for decoded data size to avoid unnecessary reallocs\r
- bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component\r
- raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;\r
- z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);\r
- if (z->expanded == NULL) return 0; // zlib should set error\r
- STBI_FREE(z->idata); z->idata = NULL;\r
- if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans)\r
- s->img_out_n = s->img_n+1;\r
- else\r
- s->img_out_n = s->img_n;\r
- if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0;\r
- if (has_trans) {\r
- if (z->depth == 16) {\r
- if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0;\r
- } else {\r
- if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;\r
- }\r
- }\r
- if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)\r
- stbi__de_iphone(z);\r
- if (pal_img_n) {\r
- // pal_img_n == 3 or 4\r
- s->img_n = pal_img_n; // record the actual colors we had\r
- s->img_out_n = pal_img_n;\r
- if (req_comp >= 3) s->img_out_n = req_comp;\r
- if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))\r
- return 0;\r
- } else if (has_trans) {\r
- // non-paletted image with tRNS -> source image has (constant) alpha\r
- ++s->img_n;\r
- }\r
- STBI_FREE(z->expanded); z->expanded = NULL;\r
- return 1;\r
- }\r
-\r
- default:\r
- // if critical, fail\r
- if (first) return stbi__err("first not IHDR", "Corrupt PNG");\r
- if ((c.type & (1 << 29)) == 0) {\r
- #ifndef STBI_NO_FAILURE_STRINGS\r
- // not threadsafe\r
- static char invalid_chunk[] = "XXXX PNG chunk not known";\r
- invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);\r
- invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);\r
- invalid_chunk[2] = STBI__BYTECAST(c.type >> 8);\r
- invalid_chunk[3] = STBI__BYTECAST(c.type >> 0);\r
- #endif\r
- return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type");\r
- }\r
- stbi__skip(s, c.length);\r
- break;\r
- }\r
- // end of PNG chunk, read and skip CRC\r
- stbi__get32be(s);\r
- }\r
-}\r
-\r
-static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri)\r
-{\r
- void *result=NULL;\r
- if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");\r
- if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) {\r
- if (p->depth < 8)\r
- ri->bits_per_channel = 8;\r
- else\r
- ri->bits_per_channel = p->depth;\r
- result = p->out;\r
- p->out = NULL;\r
- if (req_comp && req_comp != p->s->img_out_n) {\r
- if (ri->bits_per_channel == 8)\r
- result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);\r
- else\r
- result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);\r
- p->s->img_out_n = req_comp;\r
- if (result == NULL) return result;\r
- }\r
- *x = p->s->img_x;\r
- *y = p->s->img_y;\r
- if (n) *n = p->s->img_n;\r
- }\r
- STBI_FREE(p->out); p->out = NULL;\r
- STBI_FREE(p->expanded); p->expanded = NULL;\r
- STBI_FREE(p->idata); p->idata = NULL;\r
-\r
- return result;\r
-}\r
-\r
-static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- stbi__png p;\r
- p.s = s;\r
- return stbi__do_png(&p, x,y,comp,req_comp, ri);\r
-}\r
-\r
-static int stbi__png_test(stbi__context *s)\r
-{\r
- int r;\r
- r = stbi__check_png_header(s);\r
- stbi__rewind(s);\r
- return r;\r
-}\r
-\r
-static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp)\r
-{\r
- if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) {\r
- stbi__rewind( p->s );\r
- return 0;\r
- }\r
- if (x) *x = p->s->img_x;\r
- if (y) *y = p->s->img_y;\r
- if (comp) *comp = p->s->img_n;\r
- return 1;\r
-}\r
-\r
-static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- stbi__png p;\r
- p.s = s;\r
- return stbi__png_info_raw(&p, x, y, comp);\r
-}\r
-\r
-static int stbi__png_is16(stbi__context *s)\r
-{\r
- stbi__png p;\r
- p.s = s;\r
- if (!stbi__png_info_raw(&p, NULL, NULL, NULL))\r
- return 0;\r
- if (p.depth != 16) {\r
- stbi__rewind(p.s);\r
- return 0;\r
- }\r
- return 1;\r
-}\r
-#endif\r
-\r
-// Microsoft/Windows BMP image\r
-\r
-#ifndef STBI_NO_BMP\r
-static int stbi__bmp_test_raw(stbi__context *s)\r
-{\r
- int r;\r
- int sz;\r
- if (stbi__get8(s) != 'B') return 0;\r
- if (stbi__get8(s) != 'M') return 0;\r
- stbi__get32le(s); // discard filesize\r
- stbi__get16le(s); // discard reserved\r
- stbi__get16le(s); // discard reserved\r
- stbi__get32le(s); // discard data offset\r
- sz = stbi__get32le(s);\r
- r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124);\r
- return r;\r
-}\r
-\r
-static int stbi__bmp_test(stbi__context *s)\r
-{\r
- int r = stbi__bmp_test_raw(s);\r
- stbi__rewind(s);\r
- return r;\r
-}\r
-\r
-\r
-// returns 0..31 for the highest set bit\r
-static int stbi__high_bit(unsigned int z)\r
-{\r
- int n=0;\r
- if (z == 0) return -1;\r
- if (z >= 0x10000) { n += 16; z >>= 16; }\r
- if (z >= 0x00100) { n += 8; z >>= 8; }\r
- if (z >= 0x00010) { n += 4; z >>= 4; }\r
- if (z >= 0x00004) { n += 2; z >>= 2; }\r
- if (z >= 0x00002) { n += 1;/* >>= 1;*/ }\r
- return n;\r
-}\r
-\r
-static int stbi__bitcount(unsigned int a)\r
-{\r
- a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2\r
- a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4\r
- a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits\r
- a = (a + (a >> 8)); // max 16 per 8 bits\r
- a = (a + (a >> 16)); // max 32 per 8 bits\r
- return a & 0xff;\r
-}\r
-\r
-// extract an arbitrarily-aligned N-bit value (N=bits)\r
-// from v, and then make it 8-bits long and fractionally\r
-// extend it to full full range.\r
-static int stbi__shiftsigned(unsigned int v, int shift, int bits)\r
-{\r
- static unsigned int mul_table[9] = {\r
- 0,\r
- 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/,\r
- 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/,\r
- };\r
- static unsigned int shift_table[9] = {\r
- 0, 0,0,1,0,2,4,6,0,\r
- };\r
- if (shift < 0)\r
- v <<= -shift;\r
- else\r
- v >>= shift;\r
- STBI_ASSERT(v >= 0 && v < 256);\r
- v >>= (8-bits);\r
- STBI_ASSERT(bits >= 0 && bits <= 8);\r
- return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits];\r
-}\r
-\r
-typedef struct\r
-{\r
- int bpp, offset, hsz;\r
- unsigned int mr,mg,mb,ma, all_a;\r
-} stbi__bmp_data;\r
-\r
-static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info)\r
-{\r
- int hsz;\r
- if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP");\r
- stbi__get32le(s); // discard filesize\r
- stbi__get16le(s); // discard reserved\r
- stbi__get16le(s); // discard reserved\r
- info->offset = stbi__get32le(s);\r
- info->hsz = hsz = stbi__get32le(s);\r
- info->mr = info->mg = info->mb = info->ma = 0;\r
-\r
- if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown");\r
- if (hsz == 12) {\r
- s->img_x = stbi__get16le(s);\r
- s->img_y = stbi__get16le(s);\r
- } else {\r
- s->img_x = stbi__get32le(s);\r
- s->img_y = stbi__get32le(s);\r
- }\r
- if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP");\r
- info->bpp = stbi__get16le(s);\r
- if (hsz != 12) {\r
- int compress = stbi__get32le(s);\r
- if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE");\r
- stbi__get32le(s); // discard sizeof\r
- stbi__get32le(s); // discard hres\r
- stbi__get32le(s); // discard vres\r
- stbi__get32le(s); // discard colorsused\r
- stbi__get32le(s); // discard max important\r
- if (hsz == 40 || hsz == 56) {\r
- if (hsz == 56) {\r
- stbi__get32le(s);\r
- stbi__get32le(s);\r
- stbi__get32le(s);\r
- stbi__get32le(s);\r
- }\r
- if (info->bpp == 16 || info->bpp == 32) {\r
- if (compress == 0) {\r
- if (info->bpp == 32) {\r
- info->mr = 0xffu << 16;\r
- info->mg = 0xffu << 8;\r
- info->mb = 0xffu << 0;\r
- info->ma = 0xffu << 24;\r
- info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0\r
- } else {\r
- info->mr = 31u << 10;\r
- info->mg = 31u << 5;\r
- info->mb = 31u << 0;\r
- }\r
- } else if (compress == 3) {\r
- info->mr = stbi__get32le(s);\r
- info->mg = stbi__get32le(s);\r
- info->mb = stbi__get32le(s);\r
- // not documented, but generated by photoshop and handled by mspaint\r
- if (info->mr == info->mg && info->mg == info->mb) {\r
- // ?!?!?\r
- return stbi__errpuc("bad BMP", "bad BMP");\r
- }\r
- } else\r
- return stbi__errpuc("bad BMP", "bad BMP");\r
- }\r
- } else {\r
- int i;\r
- if (hsz != 108 && hsz != 124)\r
- return stbi__errpuc("bad BMP", "bad BMP");\r
- info->mr = stbi__get32le(s);\r
- info->mg = stbi__get32le(s);\r
- info->mb = stbi__get32le(s);\r
- info->ma = stbi__get32le(s);\r
- stbi__get32le(s); // discard color space\r
- for (i=0; i < 12; ++i)\r
- stbi__get32le(s); // discard color space parameters\r
- if (hsz == 124) {\r
- stbi__get32le(s); // discard rendering intent\r
- stbi__get32le(s); // discard offset of profile data\r
- stbi__get32le(s); // discard size of profile data\r
- stbi__get32le(s); // discard reserved\r
- }\r
- }\r
- }\r
- return (void *) 1;\r
-}\r
-\r
-\r
-static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- stbi_uc *out;\r
- unsigned int mr=0,mg=0,mb=0,ma=0, all_a;\r
- stbi_uc pal[256][4];\r
- int psize=0,i,j,width;\r
- int flip_vertically, pad, target;\r
- stbi__bmp_data info;\r
- STBI_NOTUSED(ri);\r
-\r
- info.all_a = 255;\r
- if (stbi__bmp_parse_header(s, &info) == NULL)\r
- return NULL; // error code already set\r
-\r
- flip_vertically = ((int) s->img_y) > 0;\r
- s->img_y = abs((int) s->img_y);\r
-\r
- mr = info.mr;\r
- mg = info.mg;\r
- mb = info.mb;\r
- ma = info.ma;\r
- all_a = info.all_a;\r
-\r
- if (info.hsz == 12) {\r
- if (info.bpp < 24)\r
- psize = (info.offset - 14 - 24) / 3;\r
- } else {\r
- if (info.bpp < 16)\r
- psize = (info.offset - 14 - info.hsz) >> 2;\r
- }\r
-\r
- if (info.bpp == 24 && ma == 0xff000000)\r
- s->img_n = 3;\r
- else\r
- s->img_n = ma ? 4 : 3;\r
- if (req_comp && req_comp >= 3) // we can directly decode 3 or 4\r
- target = req_comp;\r
- else\r
- target = s->img_n; // if they want monochrome, we'll post-convert\r
-\r
- // sanity-check size\r
- if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0))\r
- return stbi__errpuc("too large", "Corrupt BMP");\r
-\r
- out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0);\r
- if (!out) return stbi__errpuc("outofmem", "Out of memory");\r
- if (info.bpp < 16) {\r
- int z=0;\r
- if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); }\r
- for (i=0; i < psize; ++i) {\r
- pal[i][2] = stbi__get8(s);\r
- pal[i][1] = stbi__get8(s);\r
- pal[i][0] = stbi__get8(s);\r
- if (info.hsz != 12) stbi__get8(s);\r
- pal[i][3] = 255;\r
- }\r
- stbi__skip(s, info.offset - 14 - info.hsz - psize * (info.hsz == 12 ? 3 : 4));\r
- if (info.bpp == 1) width = (s->img_x + 7) >> 3;\r
- else if (info.bpp == 4) width = (s->img_x + 1) >> 1;\r
- else if (info.bpp == 8) width = s->img_x;\r
- else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); }\r
- pad = (-width)&3;\r
- if (info.bpp == 1) {\r
- for (j=0; j < (int) s->img_y; ++j) {\r
- int bit_offset = 7, v = stbi__get8(s);\r
- for (i=0; i < (int) s->img_x; ++i) {\r
- int color = (v>>bit_offset)&0x1;\r
- out[z++] = pal[color][0];\r
- out[z++] = pal[color][1];\r
- out[z++] = pal[color][2];\r
- if (target == 4) out[z++] = 255;\r
- if (i+1 == (int) s->img_x) break;\r
- if((--bit_offset) < 0) {\r
- bit_offset = 7;\r
- v = stbi__get8(s);\r
- }\r
- }\r
- stbi__skip(s, pad);\r
- }\r
- } else {\r
- for (j=0; j < (int) s->img_y; ++j) {\r
- for (i=0; i < (int) s->img_x; i += 2) {\r
- int v=stbi__get8(s),v2=0;\r
- if (info.bpp == 4) {\r
- v2 = v & 15;\r
- v >>= 4;\r
- }\r
- out[z++] = pal[v][0];\r
- out[z++] = pal[v][1];\r
- out[z++] = pal[v][2];\r
- if (target == 4) out[z++] = 255;\r
- if (i+1 == (int) s->img_x) break;\r
- v = (info.bpp == 8) ? stbi__get8(s) : v2;\r
- out[z++] = pal[v][0];\r
- out[z++] = pal[v][1];\r
- out[z++] = pal[v][2];\r
- if (target == 4) out[z++] = 255;\r
- }\r
- stbi__skip(s, pad);\r
- }\r
- }\r
- } else {\r
- int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0;\r
- int z = 0;\r
- int easy=0;\r
- stbi__skip(s, info.offset - 14 - info.hsz);\r
- if (info.bpp == 24) width = 3 * s->img_x;\r
- else if (info.bpp == 16) width = 2*s->img_x;\r
- else /* bpp = 32 and pad = 0 */ width=0;\r
- pad = (-width) & 3;\r
- if (info.bpp == 24) {\r
- easy = 1;\r
- } else if (info.bpp == 32) {\r
- if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000)\r
- easy = 2;\r
- }\r
- if (!easy) {\r
- if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); }\r
- // right shift amt to put high bit in position #7\r
- rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr);\r
- gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg);\r
- bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb);\r
- ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma);\r
- }\r
- for (j=0; j < (int) s->img_y; ++j) {\r
- if (easy) {\r
- for (i=0; i < (int) s->img_x; ++i) {\r
- unsigned char a;\r
- out[z+2] = stbi__get8(s);\r
- out[z+1] = stbi__get8(s);\r
- out[z+0] = stbi__get8(s);\r
- z += 3;\r
- a = (easy == 2 ? stbi__get8(s) : 255);\r
- all_a |= a;\r
- if (target == 4) out[z++] = a;\r
- }\r
- } else {\r
- int bpp = info.bpp;\r
- for (i=0; i < (int) s->img_x; ++i) {\r
- stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s));\r
- unsigned int a;\r
- out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount));\r
- out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount));\r
- out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount));\r
- a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255);\r
- all_a |= a;\r
- if (target == 4) out[z++] = STBI__BYTECAST(a);\r
- }\r
- }\r
- stbi__skip(s, pad);\r
- }\r
- }\r
-\r
- // if alpha channel is all 0s, replace with all 255s\r
- if (target == 4 && all_a == 0)\r
- for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4)\r
- out[i] = 255;\r
-\r
- if (flip_vertically) {\r
- stbi_uc t;\r
- for (j=0; j < (int) s->img_y>>1; ++j) {\r
- stbi_uc *p1 = out + j *s->img_x*target;\r
- stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target;\r
- for (i=0; i < (int) s->img_x*target; ++i) {\r
- t = p1[i]; p1[i] = p2[i]; p2[i] = t;\r
- }\r
- }\r
- }\r
-\r
- if (req_comp && req_comp != target) {\r
- out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y);\r
- if (out == NULL) return out; // stbi__convert_format frees input on failure\r
- }\r
-\r
- *x = s->img_x;\r
- *y = s->img_y;\r
- if (comp) *comp = s->img_n;\r
- return out;\r
-}\r
-#endif\r
-\r
-// Targa Truevision - TGA\r
-// by Jonathan Dummer\r
-#ifndef STBI_NO_TGA\r
-// returns STBI_rgb or whatever, 0 on error\r
-static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16)\r
-{\r
- // only RGB or RGBA (incl. 16bit) or grey allowed\r
- if (is_rgb16) *is_rgb16 = 0;\r
- switch(bits_per_pixel) {\r
- case 8: return STBI_grey;\r
- case 16: if(is_grey) return STBI_grey_alpha;\r
- // fallthrough\r
- case 15: if(is_rgb16) *is_rgb16 = 1;\r
- return STBI_rgb;\r
- case 24: // fallthrough\r
- case 32: return bits_per_pixel/8;\r
- default: return 0;\r
- }\r
-}\r
-\r
-static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp;\r
- int sz, tga_colormap_type;\r
- stbi__get8(s); // discard Offset\r
- tga_colormap_type = stbi__get8(s); // colormap type\r
- if( tga_colormap_type > 1 ) {\r
- stbi__rewind(s);\r
- return 0; // only RGB or indexed allowed\r
- }\r
- tga_image_type = stbi__get8(s); // image type\r
- if ( tga_colormap_type == 1 ) { // colormapped (paletted) image\r
- if (tga_image_type != 1 && tga_image_type != 9) {\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
- stbi__skip(s,4); // skip index of first colormap entry and number of entries\r
- sz = stbi__get8(s); // check bits per palette color entry\r
- if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) {\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
- stbi__skip(s,4); // skip image x and y origin\r
- tga_colormap_bpp = sz;\r
- } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE\r
- if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) {\r
- stbi__rewind(s);\r
- return 0; // only RGB or grey allowed, +/- RLE\r
- }\r
- stbi__skip(s,9); // skip colormap specification and image x/y origin\r
- tga_colormap_bpp = 0;\r
- }\r
- tga_w = stbi__get16le(s);\r
- if( tga_w < 1 ) {\r
- stbi__rewind(s);\r
- return 0; // test width\r
- }\r
- tga_h = stbi__get16le(s);\r
- if( tga_h < 1 ) {\r
- stbi__rewind(s);\r
- return 0; // test height\r
- }\r
- tga_bits_per_pixel = stbi__get8(s); // bits per pixel\r
- stbi__get8(s); // ignore alpha bits\r
- if (tga_colormap_bpp != 0) {\r
- if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) {\r
- // when using a colormap, tga_bits_per_pixel is the size of the indexes\r
- // I don't think anything but 8 or 16bit indexes makes sense\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
- tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL);\r
- } else {\r
- tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL);\r
- }\r
- if(!tga_comp) {\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
- if (x) *x = tga_w;\r
- if (y) *y = tga_h;\r
- if (comp) *comp = tga_comp;\r
- return 1; // seems to have passed everything\r
-}\r
-\r
-static int stbi__tga_test(stbi__context *s)\r
-{\r
- int res = 0;\r
- int sz, tga_color_type;\r
- stbi__get8(s); // discard Offset\r
- tga_color_type = stbi__get8(s); // color type\r
- if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed\r
- sz = stbi__get8(s); // image type\r
- if ( tga_color_type == 1 ) { // colormapped (paletted) image\r
- if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9\r
- stbi__skip(s,4); // skip index of first colormap entry and number of entries\r
- sz = stbi__get8(s); // check bits per palette color entry\r
- if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;\r
- stbi__skip(s,4); // skip image x and y origin\r
- } else { // "normal" image w/o colormap\r
- if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE\r
- stbi__skip(s,9); // skip colormap specification and image x/y origin\r
- }\r
- if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width\r
- if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height\r
- sz = stbi__get8(s); // bits per pixel\r
- if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index\r
- if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;\r
-\r
- res = 1; // if we got this far, everything's good and we can return 1 instead of 0\r
-\r
-errorEnd:\r
- stbi__rewind(s);\r
- return res;\r
-}\r
-\r
-// read 16bit value and convert to 24bit RGB\r
-static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out)\r
-{\r
- stbi__uint16 px = (stbi__uint16)stbi__get16le(s);\r
- stbi__uint16 fiveBitMask = 31;\r
- // we have 3 channels with 5bits each\r
- int r = (px >> 10) & fiveBitMask;\r
- int g = (px >> 5) & fiveBitMask;\r
- int b = px & fiveBitMask;\r
- // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later\r
- out[0] = (stbi_uc)((r * 255)/31);\r
- out[1] = (stbi_uc)((g * 255)/31);\r
- out[2] = (stbi_uc)((b * 255)/31);\r
-\r
- // some people claim that the most significant bit might be used for alpha\r
- // (possibly if an alpha-bit is set in the "image descriptor byte")\r
- // but that only made 16bit test images completely translucent..\r
- // so let's treat all 15 and 16bit TGAs as RGB with no alpha.\r
-}\r
-\r
-static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- // read in the TGA header stuff\r
- int tga_offset = stbi__get8(s);\r
- int tga_indexed = stbi__get8(s);\r
- int tga_image_type = stbi__get8(s);\r
- int tga_is_RLE = 0;\r
- int tga_palette_start = stbi__get16le(s);\r
- int tga_palette_len = stbi__get16le(s);\r
- int tga_palette_bits = stbi__get8(s);\r
- int tga_x_origin = stbi__get16le(s);\r
- int tga_y_origin = stbi__get16le(s);\r
- int tga_width = stbi__get16le(s);\r
- int tga_height = stbi__get16le(s);\r
- int tga_bits_per_pixel = stbi__get8(s);\r
- int tga_comp, tga_rgb16=0;\r
- int tga_inverted = stbi__get8(s);\r
- // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?)\r
- // image data\r
- unsigned char *tga_data;\r
- unsigned char *tga_palette = NULL;\r
- int i, j;\r
- unsigned char raw_data[4] = {0};\r
- int RLE_count = 0;\r
- int RLE_repeating = 0;\r
- int read_next_pixel = 1;\r
- STBI_NOTUSED(ri);\r
- STBI_NOTUSED(tga_x_origin); // @TODO\r
- STBI_NOTUSED(tga_y_origin); // @TODO\r
-\r
- // do a tiny bit of precessing\r
- if ( tga_image_type >= 8 )\r
- {\r
- tga_image_type -= 8;\r
- tga_is_RLE = 1;\r
- }\r
- tga_inverted = 1 - ((tga_inverted >> 5) & 1);\r
-\r
- // If I'm paletted, then I'll use the number of bits from the palette\r
- if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16);\r
- else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16);\r
-\r
- if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency\r
- return stbi__errpuc("bad format", "Can't find out TGA pixelformat");\r
-\r
- // tga info\r
- *x = tga_width;\r
- *y = tga_height;\r
- if (comp) *comp = tga_comp;\r
-\r
- if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0))\r
- return stbi__errpuc("too large", "Corrupt TGA");\r
-\r
- tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0);\r
- if (!tga_data) return stbi__errpuc("outofmem", "Out of memory");\r
-\r
- // skip to the data's starting position (offset usually = 0)\r
- stbi__skip(s, tga_offset );\r
-\r
- if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) {\r
- for (i=0; i < tga_height; ++i) {\r
- int row = tga_inverted ? tga_height -i - 1 : i;\r
- stbi_uc *tga_row = tga_data + row*tga_width*tga_comp;\r
- stbi__getn(s, tga_row, tga_width * tga_comp);\r
- }\r
- } else {\r
- // do I need to load a palette?\r
- if ( tga_indexed)\r
- {\r
- // any data to skip? (offset usually = 0)\r
- stbi__skip(s, tga_palette_start );\r
- // load the palette\r
- tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0);\r
- if (!tga_palette) {\r
- STBI_FREE(tga_data);\r
- return stbi__errpuc("outofmem", "Out of memory");\r
- }\r
- if (tga_rgb16) {\r
- stbi_uc *pal_entry = tga_palette;\r
- STBI_ASSERT(tga_comp == STBI_rgb);\r
- for (i=0; i < tga_palette_len; ++i) {\r
- stbi__tga_read_rgb16(s, pal_entry);\r
- pal_entry += tga_comp;\r
- }\r
- } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) {\r
- STBI_FREE(tga_data);\r
- STBI_FREE(tga_palette);\r
- return stbi__errpuc("bad palette", "Corrupt TGA");\r
- }\r
- }\r
- // load the data\r
- for (i=0; i < tga_width * tga_height; ++i)\r
- {\r
- // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk?\r
- if ( tga_is_RLE )\r
- {\r
- if ( RLE_count == 0 )\r
- {\r
- // yep, get the next byte as a RLE command\r
- int RLE_cmd = stbi__get8(s);\r
- RLE_count = 1 + (RLE_cmd & 127);\r
- RLE_repeating = RLE_cmd >> 7;\r
- read_next_pixel = 1;\r
- } else if ( !RLE_repeating )\r
- {\r
- read_next_pixel = 1;\r
- }\r
- } else\r
- {\r
- read_next_pixel = 1;\r
- }\r
- // OK, if I need to read a pixel, do it now\r
- if ( read_next_pixel )\r
- {\r
- // load however much data we did have\r
- if ( tga_indexed )\r
- {\r
- // read in index, then perform the lookup\r
- int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s);\r
- if ( pal_idx >= tga_palette_len ) {\r
- // invalid index\r
- pal_idx = 0;\r
- }\r
- pal_idx *= tga_comp;\r
- for (j = 0; j < tga_comp; ++j) {\r
- raw_data[j] = tga_palette[pal_idx+j];\r
- }\r
- } else if(tga_rgb16) {\r
- STBI_ASSERT(tga_comp == STBI_rgb);\r
- stbi__tga_read_rgb16(s, raw_data);\r
- } else {\r
- // read in the data raw\r
- for (j = 0; j < tga_comp; ++j) {\r
- raw_data[j] = stbi__get8(s);\r
- }\r
- }\r
- // clear the reading flag for the next pixel\r
- read_next_pixel = 0;\r
- } // end of reading a pixel\r
-\r
- // copy data\r
- for (j = 0; j < tga_comp; ++j)\r
- tga_data[i*tga_comp+j] = raw_data[j];\r
-\r
- // in case we're in RLE mode, keep counting down\r
- --RLE_count;\r
- }\r
- // do I need to invert the image?\r
- if ( tga_inverted )\r
- {\r
- for (j = 0; j*2 < tga_height; ++j)\r
- {\r
- int index1 = j * tga_width * tga_comp;\r
- int index2 = (tga_height - 1 - j) * tga_width * tga_comp;\r
- for (i = tga_width * tga_comp; i > 0; --i)\r
- {\r
- unsigned char temp = tga_data[index1];\r
- tga_data[index1] = tga_data[index2];\r
- tga_data[index2] = temp;\r
- ++index1;\r
- ++index2;\r
- }\r
- }\r
- }\r
- // clear my palette, if I had one\r
- if ( tga_palette != NULL )\r
- {\r
- STBI_FREE( tga_palette );\r
- }\r
- }\r
-\r
- // swap RGB - if the source data was RGB16, it already is in the right order\r
- if (tga_comp >= 3 && !tga_rgb16)\r
- {\r
- unsigned char* tga_pixel = tga_data;\r
- for (i=0; i < tga_width * tga_height; ++i)\r
- {\r
- unsigned char temp = tga_pixel[0];\r
- tga_pixel[0] = tga_pixel[2];\r
- tga_pixel[2] = temp;\r
- tga_pixel += tga_comp;\r
- }\r
- }\r
-\r
- // convert to target component count\r
- if (req_comp && req_comp != tga_comp)\r
- tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height);\r
-\r
- // the things I do to get rid of an error message, and yet keep\r
- // Microsoft's C compilers happy... [8^(\r
- tga_palette_start = tga_palette_len = tga_palette_bits =\r
- tga_x_origin = tga_y_origin = 0;\r
- STBI_NOTUSED(tga_palette_start);\r
- // OK, done\r
- return tga_data;\r
-}\r
-#endif\r
-\r
-// *************************************************************************************************\r
-// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB\r
-\r
-#ifndef STBI_NO_PSD\r
-static int stbi__psd_test(stbi__context *s)\r
-{\r
- int r = (stbi__get32be(s) == 0x38425053);\r
- stbi__rewind(s);\r
- return r;\r
-}\r
-\r
-static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount)\r
-{\r
- int count, nleft, len;\r
-\r
- count = 0;\r
- while ((nleft = pixelCount - count) > 0) {\r
- len = stbi__get8(s);\r
- if (len == 128) {\r
- // No-op.\r
- } else if (len < 128) {\r
- // Copy next len+1 bytes literally.\r
- len++;\r
- if (len > nleft) return 0; // corrupt data\r
- count += len;\r
- while (len) {\r
- *p = stbi__get8(s);\r
- p += 4;\r
- len--;\r
- }\r
- } else if (len > 128) {\r
- stbi_uc val;\r
- // Next -len+1 bytes in the dest are replicated from next source byte.\r
- // (Interpret len as a negative 8-bit int.)\r
- len = 257 - len;\r
- if (len > nleft) return 0; // corrupt data\r
- val = stbi__get8(s);\r
- count += len;\r
- while (len) {\r
- *p = val;\r
- p += 4;\r
- len--;\r
- }\r
- }\r
- }\r
-\r
- return 1;\r
-}\r
-\r
-static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)\r
-{\r
- int pixelCount;\r
- int channelCount, compression;\r
- int channel, i;\r
- int bitdepth;\r
- int w,h;\r
- stbi_uc *out;\r
- STBI_NOTUSED(ri);\r
-\r
- // Check identifier\r
- if (stbi__get32be(s) != 0x38425053) // "8BPS"\r
- return stbi__errpuc("not PSD", "Corrupt PSD image");\r
-\r
- // Check file type version.\r
- if (stbi__get16be(s) != 1)\r
- return stbi__errpuc("wrong version", "Unsupported version of PSD image");\r
-\r
- // Skip 6 reserved bytes.\r
- stbi__skip(s, 6 );\r
-\r
- // Read the number of channels (R, G, B, A, etc).\r
- channelCount = stbi__get16be(s);\r
- if (channelCount < 0 || channelCount > 16)\r
- return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image");\r
-\r
- // Read the rows and columns of the image.\r
- h = stbi__get32be(s);\r
- w = stbi__get32be(s);\r
-\r
- // Make sure the depth is 8 bits.\r
- bitdepth = stbi__get16be(s);\r
- if (bitdepth != 8 && bitdepth != 16)\r
- return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit");\r
-\r
- // Make sure the color mode is RGB.\r
- // Valid options are:\r
- // 0: Bitmap\r
- // 1: Grayscale\r
- // 2: Indexed color\r
- // 3: RGB color\r
- // 4: CMYK color\r
- // 7: Multichannel\r
- // 8: Duotone\r
- // 9: Lab color\r
- if (stbi__get16be(s) != 3)\r
- return stbi__errpuc("wrong color format", "PSD is not in RGB color format");\r
-\r
- // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.)\r
- stbi__skip(s,stbi__get32be(s) );\r
-\r
- // Skip the image resources. (resolution, pen tool paths, etc)\r
- stbi__skip(s, stbi__get32be(s) );\r
-\r
- // Skip the reserved data.\r
- stbi__skip(s, stbi__get32be(s) );\r
-\r
- // Find out if the data is compressed.\r
- // Known values:\r
- // 0: no compression\r
- // 1: RLE compressed\r
- compression = stbi__get16be(s);\r
- if (compression > 1)\r
- return stbi__errpuc("bad compression", "PSD has an unknown compression format");\r
-\r
- // Check size\r
- if (!stbi__mad3sizes_valid(4, w, h, 0))\r
- return stbi__errpuc("too large", "Corrupt PSD");\r
-\r
- // Create the destination image.\r
-\r
- if (!compression && bitdepth == 16 && bpc == 16) {\r
- out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0);\r
- ri->bits_per_channel = 16;\r
- } else\r
- out = (stbi_uc *) stbi__malloc(4 * w*h);\r
-\r
- if (!out) return stbi__errpuc("outofmem", "Out of memory");\r
- pixelCount = w*h;\r
-\r
- // Initialize the data to zero.\r
- //memset( out, 0, pixelCount * 4 );\r
-\r
- // Finally, the image data.\r
- if (compression) {\r
- // RLE as used by .PSD and .TIFF\r
- // Loop until you get the number of unpacked bytes you are expecting:\r
- // Read the next source byte into n.\r
- // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally.\r
- // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times.\r
- // Else if n is 128, noop.\r
- // Endloop\r
-\r
- // The RLE-compressed data is preceded by a 2-byte data count for each row in the data,\r
- // which we're going to just skip.\r
- stbi__skip(s, h * channelCount * 2 );\r
-\r
- // Read the RLE data by channel.\r
- for (channel = 0; channel < 4; channel++) {\r
- stbi_uc *p;\r
-\r
- p = out+channel;\r
- if (channel >= channelCount) {\r
- // Fill this channel with default data.\r
- for (i = 0; i < pixelCount; i++, p += 4)\r
- *p = (channel == 3 ? 255 : 0);\r
- } else {\r
- // Read the RLE data.\r
- if (!stbi__psd_decode_rle(s, p, pixelCount)) {\r
- STBI_FREE(out);\r
- return stbi__errpuc("corrupt", "bad RLE data");\r
- }\r
- }\r
- }\r
-\r
- } else {\r
- // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...)\r
- // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image.\r
-\r
- // Read the data by channel.\r
- for (channel = 0; channel < 4; channel++) {\r
- if (channel >= channelCount) {\r
- // Fill this channel with default data.\r
- if (bitdepth == 16 && bpc == 16) {\r
- stbi__uint16 *q = ((stbi__uint16 *) out) + channel;\r
- stbi__uint16 val = channel == 3 ? 65535 : 0;\r
- for (i = 0; i < pixelCount; i++, q += 4)\r
- *q = val;\r
- } else {\r
- stbi_uc *p = out+channel;\r
- stbi_uc val = channel == 3 ? 255 : 0;\r
- for (i = 0; i < pixelCount; i++, p += 4)\r
- *p = val;\r
- }\r
- } else {\r
- if (ri->bits_per_channel == 16) { // output bpc\r
- stbi__uint16 *q = ((stbi__uint16 *) out) + channel;\r
- for (i = 0; i < pixelCount; i++, q += 4)\r
- *q = (stbi__uint16) stbi__get16be(s);\r
- } else {\r
- stbi_uc *p = out+channel;\r
- if (bitdepth == 16) { // input bpc\r
- for (i = 0; i < pixelCount; i++, p += 4)\r
- *p = (stbi_uc) (stbi__get16be(s) >> 8);\r
- } else {\r
- for (i = 0; i < pixelCount; i++, p += 4)\r
- *p = stbi__get8(s);\r
- }\r
- }\r
- }\r
- }\r
- }\r
-\r
- // remove weird white matte from PSD\r
- if (channelCount >= 4) {\r
- if (ri->bits_per_channel == 16) {\r
- for (i=0; i < w*h; ++i) {\r
- stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i;\r
- if (pixel[3] != 0 && pixel[3] != 65535) {\r
- float a = pixel[3] / 65535.0f;\r
- float ra = 1.0f / a;\r
- float inv_a = 65535.0f * (1 - ra);\r
- pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a);\r
- pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a);\r
- pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a);\r
- }\r
- }\r
- } else {\r
- for (i=0; i < w*h; ++i) {\r
- unsigned char *pixel = out + 4*i;\r
- if (pixel[3] != 0 && pixel[3] != 255) {\r
- float a = pixel[3] / 255.0f;\r
- float ra = 1.0f / a;\r
- float inv_a = 255.0f * (1 - ra);\r
- pixel[0] = (unsigned char) (pixel[0]*ra + inv_a);\r
- pixel[1] = (unsigned char) (pixel[1]*ra + inv_a);\r
- pixel[2] = (unsigned char) (pixel[2]*ra + inv_a);\r
- }\r
- }\r
- }\r
- }\r
-\r
- // convert to desired output format\r
- if (req_comp && req_comp != 4) {\r
- if (ri->bits_per_channel == 16)\r
- out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h);\r
- else\r
- out = stbi__convert_format(out, 4, req_comp, w, h);\r
- if (out == NULL) return out; // stbi__convert_format frees input on failure\r
- }\r
-\r
- if (comp) *comp = 4;\r
- *y = h;\r
- *x = w;\r
-\r
- return out;\r
-}\r
-#endif\r
-\r
-// *************************************************************************************************\r
-// Softimage PIC loader\r
-// by Tom Seddon\r
-//\r
-// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format\r
-// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/\r
-\r
-#ifndef STBI_NO_PIC\r
-static int stbi__pic_is4(stbi__context *s,const char *str)\r
-{\r
- int i;\r
- for (i=0; i<4; ++i)\r
- if (stbi__get8(s) != (stbi_uc)str[i])\r
- return 0;\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__pic_test_core(stbi__context *s)\r
-{\r
- int i;\r
-\r
- if (!stbi__pic_is4(s,"\x53\x80\xF6\x34"))\r
- return 0;\r
-\r
- for(i=0;i<84;++i)\r
- stbi__get8(s);\r
-\r
- if (!stbi__pic_is4(s,"PICT"))\r
- return 0;\r
-\r
- return 1;\r
-}\r
-\r
-typedef struct\r
-{\r
- stbi_uc size,type,channel;\r
-} stbi__pic_packet;\r
-\r
-static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest)\r
-{\r
- int mask=0x80, i;\r
-\r
- for (i=0; i<4; ++i, mask>>=1) {\r
- if (channel & mask) {\r
- if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short");\r
- dest[i]=stbi__get8(s);\r
- }\r
- }\r
-\r
- return dest;\r
-}\r
-\r
-static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src)\r
-{\r
- int mask=0x80,i;\r
-\r
- for (i=0;i<4; ++i, mask>>=1)\r
- if (channel&mask)\r
- dest[i]=src[i];\r
-}\r
-\r
-static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result)\r
-{\r
- int act_comp=0,num_packets=0,y,chained;\r
- stbi__pic_packet packets[10];\r
-\r
- // this will (should...) cater for even some bizarre stuff like having data\r
- // for the same channel in multiple packets.\r
- do {\r
- stbi__pic_packet *packet;\r
-\r
- if (num_packets==sizeof(packets)/sizeof(packets[0]))\r
- return stbi__errpuc("bad format","too many packets");\r
-\r
- packet = &packets[num_packets++];\r
-\r
- chained = stbi__get8(s);\r
- packet->size = stbi__get8(s);\r
- packet->type = stbi__get8(s);\r
- packet->channel = stbi__get8(s);\r
-\r
- act_comp |= packet->channel;\r
-\r
- if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)");\r
- if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp");\r
- } while (chained);\r
-\r
- *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel?\r
-\r
- for(y=0; y<height; ++y) {\r
- int packet_idx;\r
-\r
- for(packet_idx=0; packet_idx < num_packets; ++packet_idx) {\r
- stbi__pic_packet *packet = &packets[packet_idx];\r
- stbi_uc *dest = result+y*width*4;\r
-\r
- switch (packet->type) {\r
- default:\r
- return stbi__errpuc("bad format","packet has bad compression type");\r
-\r
- case 0: {//uncompressed\r
- int x;\r
-\r
- for(x=0;x<width;++x, dest+=4)\r
- if (!stbi__readval(s,packet->channel,dest))\r
- return 0;\r
- break;\r
- }\r
-\r
- case 1://Pure RLE\r
- {\r
- int left=width, i;\r
-\r
- while (left>0) {\r
- stbi_uc count,value[4];\r
-\r
- count=stbi__get8(s);\r
- if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)");\r
-\r
- if (count > left)\r
- count = (stbi_uc) left;\r
-\r
- if (!stbi__readval(s,packet->channel,value)) return 0;\r
-\r
- for(i=0; i<count; ++i,dest+=4)\r
- stbi__copyval(packet->channel,dest,value);\r
- left -= count;\r
- }\r
- }\r
- break;\r
-\r
- case 2: {//Mixed RLE\r
- int left=width;\r
- while (left>0) {\r
- int count = stbi__get8(s), i;\r
- if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)");\r
-\r
- if (count >= 128) { // Repeated\r
- stbi_uc value[4];\r
-\r
- if (count==128)\r
- count = stbi__get16be(s);\r
- else\r
- count -= 127;\r
- if (count > left)\r
- return stbi__errpuc("bad file","scanline overrun");\r
-\r
- if (!stbi__readval(s,packet->channel,value))\r
- return 0;\r
-\r
- for(i=0;i<count;++i, dest += 4)\r
- stbi__copyval(packet->channel,dest,value);\r
- } else { // Raw\r
- ++count;\r
- if (count>left) return stbi__errpuc("bad file","scanline overrun");\r
-\r
- for(i=0;i<count;++i, dest+=4)\r
- if (!stbi__readval(s,packet->channel,dest))\r
- return 0;\r
- }\r
- left-=count;\r
- }\r
- break;\r
- }\r
- }\r
- }\r
- }\r
-\r
- return result;\r
-}\r
-\r
-static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri)\r
-{\r
- stbi_uc *result;\r
- int i, x,y, internal_comp;\r
- STBI_NOTUSED(ri);\r
-\r
- if (!comp) comp = &internal_comp;\r
-\r
- for (i=0; i<92; ++i)\r
- stbi__get8(s);\r
-\r
- x = stbi__get16be(s);\r
- y = stbi__get16be(s);\r
- if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)");\r
- if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode");\r
-\r
- stbi__get32be(s); //skip `ratio'\r
- stbi__get16be(s); //skip `fields'\r
- stbi__get16be(s); //skip `pad'\r
-\r
- // intermediate buffer is RGBA\r
- result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0);\r
- memset(result, 0xff, x*y*4);\r
-\r
- if (!stbi__pic_load_core(s,x,y,comp, result)) {\r
- STBI_FREE(result);\r
- result=0;\r
- }\r
- *px = x;\r
- *py = y;\r
- if (req_comp == 0) req_comp = *comp;\r
- result=stbi__convert_format(result,4,req_comp,x,y);\r
-\r
- return result;\r
-}\r
-\r
-static int stbi__pic_test(stbi__context *s)\r
-{\r
- int r = stbi__pic_test_core(s);\r
- stbi__rewind(s);\r
- return r;\r
-}\r
-#endif\r
-\r
-// *************************************************************************************************\r
-// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb\r
-\r
-#ifndef STBI_NO_GIF\r
-typedef struct\r
-{\r
- stbi__int16 prefix;\r
- stbi_uc first;\r
- stbi_uc suffix;\r
-} stbi__gif_lzw;\r
-\r
-typedef struct\r
-{\r
- int w,h;\r
- stbi_uc *out; // output buffer (always 4 components)\r
- stbi_uc *background; // The current "background" as far as a gif is concerned\r
- stbi_uc *history; \r
- int flags, bgindex, ratio, transparent, eflags;\r
- stbi_uc pal[256][4];\r
- stbi_uc lpal[256][4];\r
- stbi__gif_lzw codes[8192];\r
- stbi_uc *color_table;\r
- int parse, step;\r
- int lflags;\r
- int start_x, start_y;\r
- int max_x, max_y;\r
- int cur_x, cur_y;\r
- int line_size;\r
- int delay;\r
-} stbi__gif;\r
-\r
-static int stbi__gif_test_raw(stbi__context *s)\r
-{\r
- int sz;\r
- if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0;\r
- sz = stbi__get8(s);\r
- if (sz != '9' && sz != '7') return 0;\r
- if (stbi__get8(s) != 'a') return 0;\r
- return 1;\r
-}\r
-\r
-static int stbi__gif_test(stbi__context *s)\r
-{\r
- int r = stbi__gif_test_raw(s);\r
- stbi__rewind(s);\r
- return r;\r
-}\r
-\r
-static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp)\r
-{\r
- int i;\r
- for (i=0; i < num_entries; ++i) {\r
- pal[i][2] = stbi__get8(s);\r
- pal[i][1] = stbi__get8(s);\r
- pal[i][0] = stbi__get8(s);\r
- pal[i][3] = transp == i ? 0 : 255;\r
- }\r
-}\r
-\r
-static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info)\r
-{\r
- stbi_uc version;\r
- if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8')\r
- return stbi__err("not GIF", "Corrupt GIF");\r
-\r
- version = stbi__get8(s);\r
- if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF");\r
- if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF");\r
-\r
- stbi__g_failure_reason = "";\r
- g->w = stbi__get16le(s);\r
- g->h = stbi__get16le(s);\r
- g->flags = stbi__get8(s);\r
- g->bgindex = stbi__get8(s);\r
- g->ratio = stbi__get8(s);\r
- g->transparent = -1;\r
-\r
- if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments\r
-\r
- if (is_info) return 1;\r
-\r
- if (g->flags & 0x80)\r
- stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1);\r
-\r
- return 1;\r
-}\r
-\r
-static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif));\r
- if (!stbi__gif_header(s, g, comp, 1)) {\r
- STBI_FREE(g);\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- if (x) *x = g->w;\r
- if (y) *y = g->h;\r
- STBI_FREE(g);\r
- return 1;\r
-}\r
-\r
-static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code)\r
-{\r
- stbi_uc *p, *c;\r
- int idx; \r
-\r
- // recurse to decode the prefixes, since the linked-list is backwards,\r
- // and working backwards through an interleaved image would be nasty\r
- if (g->codes[code].prefix >= 0)\r
- stbi__out_gif_code(g, g->codes[code].prefix);\r
-\r
- if (g->cur_y >= g->max_y) return;\r
-\r
- idx = g->cur_x + g->cur_y; \r
- p = &g->out[idx];\r
- g->history[idx / 4] = 1; \r
-\r
- c = &g->color_table[g->codes[code].suffix * 4];\r
- if (c[3] > 128) { // don't render transparent pixels; \r
- p[0] = c[2];\r
- p[1] = c[1];\r
- p[2] = c[0];\r
- p[3] = c[3];\r
- }\r
- g->cur_x += 4;\r
-\r
- if (g->cur_x >= g->max_x) {\r
- g->cur_x = g->start_x;\r
- g->cur_y += g->step;\r
-\r
- while (g->cur_y >= g->max_y && g->parse > 0) {\r
- g->step = (1 << g->parse) * g->line_size;\r
- g->cur_y = g->start_y + (g->step >> 1);\r
- --g->parse;\r
- }\r
- }\r
-}\r
-\r
-static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g)\r
-{\r
- stbi_uc lzw_cs;\r
- stbi__int32 len, init_code;\r
- stbi__uint32 first;\r
- stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear;\r
- stbi__gif_lzw *p;\r
-\r
- lzw_cs = stbi__get8(s);\r
- if (lzw_cs > 12) return NULL;\r
- clear = 1 << lzw_cs;\r
- first = 1;\r
- codesize = lzw_cs + 1;\r
- codemask = (1 << codesize) - 1;\r
- bits = 0;\r
- valid_bits = 0;\r
- for (init_code = 0; init_code < clear; init_code++) {\r
- g->codes[init_code].prefix = -1;\r
- g->codes[init_code].first = (stbi_uc) init_code;\r
- g->codes[init_code].suffix = (stbi_uc) init_code;\r
- }\r
-\r
- // support no starting clear code\r
- avail = clear+2;\r
- oldcode = -1;\r
-\r
- len = 0;\r
- for(;;) {\r
- if (valid_bits < codesize) {\r
- if (len == 0) {\r
- len = stbi__get8(s); // start new block\r
- if (len == 0)\r
- return g->out;\r
- }\r
- --len;\r
- bits |= (stbi__int32) stbi__get8(s) << valid_bits;\r
- valid_bits += 8;\r
- } else {\r
- stbi__int32 code = bits & codemask;\r
- bits >>= codesize;\r
- valid_bits -= codesize;\r
- // @OPTIMIZE: is there some way we can accelerate the non-clear path?\r
- if (code == clear) { // clear code\r
- codesize = lzw_cs + 1;\r
- codemask = (1 << codesize) - 1;\r
- avail = clear + 2;\r
- oldcode = -1;\r
- first = 0;\r
- } else if (code == clear + 1) { // end of stream code\r
- stbi__skip(s, len);\r
- while ((len = stbi__get8(s)) > 0)\r
- stbi__skip(s,len);\r
- return g->out;\r
- } else if (code <= avail) {\r
- if (first) {\r
- return stbi__errpuc("no clear code", "Corrupt GIF");\r
- }\r
-\r
- if (oldcode >= 0) {\r
- p = &g->codes[avail++];\r
- if (avail > 8192) {\r
- return stbi__errpuc("too many codes", "Corrupt GIF");\r
- }\r
-\r
- p->prefix = (stbi__int16) oldcode;\r
- p->first = g->codes[oldcode].first;\r
- p->suffix = (code == avail) ? p->first : g->codes[code].first;\r
- } else if (code == avail)\r
- return stbi__errpuc("illegal code in raster", "Corrupt GIF");\r
-\r
- stbi__out_gif_code(g, (stbi__uint16) code);\r
-\r
- if ((avail & codemask) == 0 && avail <= 0x0FFF) {\r
- codesize++;\r
- codemask = (1 << codesize) - 1;\r
- }\r
-\r
- oldcode = code;\r
- } else {\r
- return stbi__errpuc("illegal code in raster", "Corrupt GIF");\r
- }\r
- }\r
- }\r
-}\r
-\r
-// this function is designed to support animated gifs, although stb_image doesn't support it\r
-// two back is the image from two frames ago, used for a very specific disposal format\r
-static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back)\r
-{\r
- int dispose; \r
- int first_frame; \r
- int pi; \r
- int pcount; \r
- STBI_NOTUSED(req_comp);\r
-\r
- // on first frame, any non-written pixels get the background colour (non-transparent)\r
- first_frame = 0; \r
- if (g->out == 0) {\r
- if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header\r
- if (!stbi__mad3sizes_valid(4, g->w, g->h, 0))\r
- return stbi__errpuc("too large", "GIF image is too large");\r
- pcount = g->w * g->h;\r
- g->out = (stbi_uc *) stbi__malloc(4 * pcount);\r
- g->background = (stbi_uc *) stbi__malloc(4 * pcount);\r
- g->history = (stbi_uc *) stbi__malloc(pcount);\r
- if (!g->out || !g->background || !g->history)\r
- return stbi__errpuc("outofmem", "Out of memory");\r
-\r
- // image is treated as "transparent" at the start - ie, nothing overwrites the current background; \r
- // background colour is only used for pixels that are not rendered first frame, after that "background"\r
- // color refers to the color that was there the previous frame. \r
- memset(g->out, 0x00, 4 * pcount);\r
- memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent)\r
- memset(g->history, 0x00, pcount); // pixels that were affected previous frame\r
- first_frame = 1; \r
- } else {\r
- // second frame - how do we dispoase of the previous one?\r
- dispose = (g->eflags & 0x1C) >> 2; \r
- pcount = g->w * g->h; \r
-\r
- if ((dispose == 3) && (two_back == 0)) {\r
- dispose = 2; // if I don't have an image to revert back to, default to the old background\r
- }\r
-\r
- if (dispose == 3) { // use previous graphic\r
- for (pi = 0; pi < pcount; ++pi) {\r
- if (g->history[pi]) {\r
- memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); \r
- }\r
- }\r
- } else if (dispose == 2) { \r
- // restore what was changed last frame to background before that frame; \r
- for (pi = 0; pi < pcount; ++pi) {\r
- if (g->history[pi]) {\r
- memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); \r
- }\r
- }\r
- } else {\r
- // This is a non-disposal case eithe way, so just \r
- // leave the pixels as is, and they will become the new background\r
- // 1: do not dispose\r
- // 0: not specified.\r
- }\r
-\r
- // background is what out is after the undoing of the previou frame; \r
- memcpy( g->background, g->out, 4 * g->w * g->h ); \r
- }\r
-\r
- // clear my history; \r
- memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame\r
-\r
- for (;;) {\r
- int tag = stbi__get8(s); \r
- switch (tag) {\r
- case 0x2C: /* Image Descriptor */\r
- {\r
- stbi__int32 x, y, w, h;\r
- stbi_uc *o;\r
-\r
- x = stbi__get16le(s);\r
- y = stbi__get16le(s);\r
- w = stbi__get16le(s);\r
- h = stbi__get16le(s);\r
- if (((x + w) > (g->w)) || ((y + h) > (g->h)))\r
- return stbi__errpuc("bad Image Descriptor", "Corrupt GIF");\r
-\r
- g->line_size = g->w * 4;\r
- g->start_x = x * 4;\r
- g->start_y = y * g->line_size;\r
- g->max_x = g->start_x + w * 4;\r
- g->max_y = g->start_y + h * g->line_size;\r
- g->cur_x = g->start_x;\r
- g->cur_y = g->start_y;\r
-\r
- // if the width of the specified rectangle is 0, that means\r
- // we may not see *any* pixels or the image is malformed;\r
- // to make sure this is caught, move the current y down to\r
- // max_y (which is what out_gif_code checks).\r
- if (w == 0)\r
- g->cur_y = g->max_y;\r
-\r
- g->lflags = stbi__get8(s);\r
-\r
- if (g->lflags & 0x40) {\r
- g->step = 8 * g->line_size; // first interlaced spacing\r
- g->parse = 3;\r
- } else {\r
- g->step = g->line_size;\r
- g->parse = 0;\r
- }\r
-\r
- if (g->lflags & 0x80) {\r
- stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1);\r
- g->color_table = (stbi_uc *) g->lpal;\r
- } else if (g->flags & 0x80) {\r
- g->color_table = (stbi_uc *) g->pal;\r
- } else\r
- return stbi__errpuc("missing color table", "Corrupt GIF"); \r
- \r
- o = stbi__process_gif_raster(s, g);\r
- if (!o) return NULL;\r
-\r
- // if this was the first frame, \r
- pcount = g->w * g->h; \r
- if (first_frame && (g->bgindex > 0)) {\r
- // if first frame, any pixel not drawn to gets the background color\r
- for (pi = 0; pi < pcount; ++pi) {\r
- if (g->history[pi] == 0) {\r
- g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; \r
- memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); \r
- }\r
- }\r
- }\r
-\r
- return o;\r
- }\r
-\r
- case 0x21: // Comment Extension.\r
- {\r
- int len;\r
- int ext = stbi__get8(s); \r
- if (ext == 0xF9) { // Graphic Control Extension.\r
- len = stbi__get8(s);\r
- if (len == 4) {\r
- g->eflags = stbi__get8(s);\r
- g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths.\r
-\r
- // unset old transparent\r
- if (g->transparent >= 0) {\r
- g->pal[g->transparent][3] = 255; \r
- } \r
- if (g->eflags & 0x01) {\r
- g->transparent = stbi__get8(s);\r
- if (g->transparent >= 0) {\r
- g->pal[g->transparent][3] = 0; \r
- }\r
- } else {\r
- // don't need transparent\r
- stbi__skip(s, 1); \r
- g->transparent = -1; \r
- }\r
- } else {\r
- stbi__skip(s, len);\r
- break;\r
- }\r
- } \r
- while ((len = stbi__get8(s)) != 0) {\r
- stbi__skip(s, len);\r
- }\r
- break;\r
- }\r
-\r
- case 0x3B: // gif stream termination code\r
- return (stbi_uc *) s; // using '1' causes warning on some compilers\r
-\r
- default:\r
- return stbi__errpuc("unknown code", "Corrupt GIF");\r
- }\r
- }\r
-}\r
-\r
-static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp)\r
-{\r
- if (stbi__gif_test(s)) {\r
- int layers = 0; \r
- stbi_uc *u = 0;\r
- stbi_uc *out = 0;\r
- stbi_uc *two_back = 0; \r
- stbi__gif g;\r
- int stride; \r
- memset(&g, 0, sizeof(g));\r
- if (delays) {\r
- *delays = 0; \r
- }\r
-\r
- do {\r
- u = stbi__gif_load_next(s, &g, comp, req_comp, two_back);\r
- if (u == (stbi_uc *) s) u = 0; // end of animated gif marker\r
-\r
- if (u) {\r
- *x = g.w;\r
- *y = g.h;\r
- ++layers; \r
- stride = g.w * g.h * 4; \r
- \r
- if (out) {\r
- out = (stbi_uc*) STBI_REALLOC( out, layers * stride ); \r
- if (delays) {\r
- *delays = (int*) STBI_REALLOC( *delays, sizeof(int) * layers ); \r
- }\r
- } else {\r
- out = (stbi_uc*)stbi__malloc( layers * stride ); \r
- if (delays) {\r
- *delays = (int*) stbi__malloc( layers * sizeof(int) ); \r
- }\r
- }\r
- memcpy( out + ((layers - 1) * stride), u, stride ); \r
- if (layers >= 2) {\r
- two_back = out - 2 * stride; \r
- }\r
-\r
- if (delays) {\r
- (*delays)[layers - 1U] = g.delay; \r
- }\r
- }\r
- } while (u != 0); \r
-\r
- // free temp buffer; \r
- STBI_FREE(g.out); \r
- STBI_FREE(g.history); \r
- STBI_FREE(g.background); \r
-\r
- // do the final conversion after loading everything; \r
- if (req_comp && req_comp != 4)\r
- out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h);\r
-\r
- *z = layers; \r
- return out;\r
- } else {\r
- return stbi__errpuc("not GIF", "Image was not as a gif type."); \r
- }\r
-}\r
-\r
-static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- stbi_uc *u = 0;\r
- stbi__gif g;\r
- memset(&g, 0, sizeof(g));\r
- STBI_NOTUSED(ri);\r
-\r
- u = stbi__gif_load_next(s, &g, comp, req_comp, 0);\r
- if (u == (stbi_uc *) s) u = 0; // end of animated gif marker\r
- if (u) {\r
- *x = g.w;\r
- *y = g.h;\r
-\r
- // moved conversion to after successful load so that the same\r
- // can be done for multiple frames. \r
- if (req_comp && req_comp != 4)\r
- u = stbi__convert_format(u, 4, req_comp, g.w, g.h);\r
- } else if (g.out) {\r
- // if there was an error and we allocated an image buffer, free it!\r
- STBI_FREE(g.out);\r
- }\r
-\r
- // free buffers needed for multiple frame loading; \r
- STBI_FREE(g.history);\r
- STBI_FREE(g.background); \r
-\r
- return u;\r
-}\r
-\r
-static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- return stbi__gif_info_raw(s,x,y,comp);\r
-}\r
-#endif\r
-\r
-// *************************************************************************************************\r
-// Radiance RGBE HDR loader\r
-// originally by Nicolas Schulz\r
-#ifndef STBI_NO_HDR\r
-static int stbi__hdr_test_core(stbi__context *s, const char *signature)\r
-{\r
- int i;\r
- for (i=0; signature[i]; ++i)\r
- if (stbi__get8(s) != signature[i])\r
- return 0;\r
- stbi__rewind(s);\r
- return 1;\r
-}\r
-\r
-static int stbi__hdr_test(stbi__context* s)\r
-{\r
- int r = stbi__hdr_test_core(s, "#?RADIANCE\n");\r
- stbi__rewind(s);\r
- if(!r) {\r
- r = stbi__hdr_test_core(s, "#?RGBE\n");\r
- stbi__rewind(s);\r
- }\r
- return r;\r
-}\r
-\r
-#define STBI__HDR_BUFLEN 1024\r
-static char *stbi__hdr_gettoken(stbi__context *z, char *buffer)\r
-{\r
- int len=0;\r
- char c = '\0';\r
-\r
- c = (char) stbi__get8(z);\r
-\r
- while (!stbi__at_eof(z) && c != '\n') {\r
- buffer[len++] = c;\r
- if (len == STBI__HDR_BUFLEN-1) {\r
- // flush to end of line\r
- while (!stbi__at_eof(z) && stbi__get8(z) != '\n')\r
- ;\r
- break;\r
- }\r
- c = (char) stbi__get8(z);\r
- }\r
-\r
- buffer[len] = 0;\r
- return buffer;\r
-}\r
-\r
-static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp)\r
-{\r
- if ( input[3] != 0 ) {\r
- float f1;\r
- // Exponent\r
- f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8));\r
- if (req_comp <= 2)\r
- output[0] = (input[0] + input[1] + input[2]) * f1 / 3;\r
- else {\r
- output[0] = input[0] * f1;\r
- output[1] = input[1] * f1;\r
- output[2] = input[2] * f1;\r
- }\r
- if (req_comp == 2) output[1] = 1;\r
- if (req_comp == 4) output[3] = 1;\r
- } else {\r
- switch (req_comp) {\r
- case 4: output[3] = 1; /* fallthrough */\r
- case 3: output[0] = output[1] = output[2] = 0;\r
- break;\r
- case 2: output[1] = 1; /* fallthrough */\r
- case 1: output[0] = 0;\r
- break;\r
- }\r
- }\r
-}\r
-\r
-static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- char buffer[STBI__HDR_BUFLEN];\r
- char *token;\r
- int valid = 0;\r
- int width, height;\r
- stbi_uc *scanline;\r
- float *hdr_data;\r
- int len;\r
- unsigned char count, value;\r
- int i, j, k, c1,c2, z;\r
- const char *headerToken;\r
- STBI_NOTUSED(ri);\r
-\r
- // Check identifier\r
- headerToken = stbi__hdr_gettoken(s,buffer);\r
- if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0)\r
- return stbi__errpf("not HDR", "Corrupt HDR image");\r
-\r
- // Parse header\r
- for(;;) {\r
- token = stbi__hdr_gettoken(s,buffer);\r
- if (token[0] == 0) break;\r
- if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;\r
- }\r
-\r
- if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format");\r
-\r
- // Parse width and height\r
- // can't use sscanf() if we're not using stdio!\r
- token = stbi__hdr_gettoken(s,buffer);\r
- if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format");\r
- token += 3;\r
- height = (int) strtol(token, &token, 10);\r
- while (*token == ' ') ++token;\r
- if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format");\r
- token += 3;\r
- width = (int) strtol(token, NULL, 10);\r
-\r
- *x = width;\r
- *y = height;\r
-\r
- if (comp) *comp = 3;\r
- if (req_comp == 0) req_comp = 3;\r
-\r
- if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0))\r
- return stbi__errpf("too large", "HDR image is too large");\r
-\r
- // Read data\r
- hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0);\r
- if (!hdr_data)\r
- return stbi__errpf("outofmem", "Out of memory");\r
-\r
- // Load image data\r
- // image data is stored as some number of sca\r
- if ( width < 8 || width >= 32768) {\r
- // Read flat data\r
- for (j=0; j < height; ++j) {\r
- for (i=0; i < width; ++i) {\r
- stbi_uc rgbe[4];\r
- main_decode_loop:\r
- stbi__getn(s, rgbe, 4);\r
- stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp);\r
- }\r
- }\r
- } else {\r
- // Read RLE-encoded data\r
- scanline = NULL;\r
-\r
- for (j = 0; j < height; ++j) {\r
- c1 = stbi__get8(s);\r
- c2 = stbi__get8(s);\r
- len = stbi__get8(s);\r
- if (c1 != 2 || c2 != 2 || (len & 0x80)) {\r
- // not run-length encoded, so we have to actually use THIS data as a decoded\r
- // pixel (note this can't be a valid pixel--one of RGB must be >= 128)\r
- stbi_uc rgbe[4];\r
- rgbe[0] = (stbi_uc) c1;\r
- rgbe[1] = (stbi_uc) c2;\r
- rgbe[2] = (stbi_uc) len;\r
- rgbe[3] = (stbi_uc) stbi__get8(s);\r
- stbi__hdr_convert(hdr_data, rgbe, req_comp);\r
- i = 1;\r
- j = 0;\r
- STBI_FREE(scanline);\r
- goto main_decode_loop; // yes, this makes no sense\r
- }\r
- len <<= 8;\r
- len |= stbi__get8(s);\r
- if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); }\r
- if (scanline == NULL) {\r
- scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0);\r
- if (!scanline) {\r
- STBI_FREE(hdr_data);\r
- return stbi__errpf("outofmem", "Out of memory");\r
- }\r
- }\r
-\r
- for (k = 0; k < 4; ++k) {\r
- int nleft;\r
- i = 0;\r
- while ((nleft = width - i) > 0) {\r
- count = stbi__get8(s);\r
- if (count > 128) {\r
- // Run\r
- value = stbi__get8(s);\r
- count -= 128;\r
- if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }\r
- for (z = 0; z < count; ++z)\r
- scanline[i++ * 4 + k] = value;\r
- } else {\r
- // Dump\r
- if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }\r
- for (z = 0; z < count; ++z)\r
- scanline[i++ * 4 + k] = stbi__get8(s);\r
- }\r
- }\r
- }\r
- for (i=0; i < width; ++i)\r
- stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp);\r
- }\r
- if (scanline)\r
- STBI_FREE(scanline);\r
- }\r
-\r
- return hdr_data;\r
-}\r
-\r
-static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- char buffer[STBI__HDR_BUFLEN];\r
- char *token;\r
- int valid = 0;\r
- int dummy;\r
-\r
- if (!x) x = &dummy;\r
- if (!y) y = &dummy;\r
- if (!comp) comp = &dummy;\r
-\r
- if (stbi__hdr_test(s) == 0) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
-\r
- for(;;) {\r
- token = stbi__hdr_gettoken(s,buffer);\r
- if (token[0] == 0) break;\r
- if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;\r
- }\r
-\r
- if (!valid) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- token = stbi__hdr_gettoken(s,buffer);\r
- if (strncmp(token, "-Y ", 3)) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- token += 3;\r
- *y = (int) strtol(token, &token, 10);\r
- while (*token == ' ') ++token;\r
- if (strncmp(token, "+X ", 3)) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- token += 3;\r
- *x = (int) strtol(token, NULL, 10);\r
- *comp = 3;\r
- return 1;\r
-}\r
-#endif // STBI_NO_HDR\r
-\r
-#ifndef STBI_NO_BMP\r
-static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- void *p;\r
- stbi__bmp_data info;\r
-\r
- info.all_a = 255;\r
- p = stbi__bmp_parse_header(s, &info);\r
- stbi__rewind( s );\r
- if (p == NULL)\r
- return 0;\r
- if (x) *x = s->img_x;\r
- if (y) *y = s->img_y;\r
- if (comp) {\r
- if (info.bpp == 24 && info.ma == 0xff000000)\r
- *comp = 3;\r
- else\r
- *comp = info.ma ? 4 : 3;\r
- }\r
- return 1;\r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_PSD\r
-static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- int channelCount, dummy, depth;\r
- if (!x) x = &dummy;\r
- if (!y) y = &dummy;\r
- if (!comp) comp = &dummy;\r
- if (stbi__get32be(s) != 0x38425053) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- if (stbi__get16be(s) != 1) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- stbi__skip(s, 6);\r
- channelCount = stbi__get16be(s);\r
- if (channelCount < 0 || channelCount > 16) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- *y = stbi__get32be(s);\r
- *x = stbi__get32be(s);\r
- depth = stbi__get16be(s);\r
- if (depth != 8 && depth != 16) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- if (stbi__get16be(s) != 3) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- *comp = 4;\r
- return 1;\r
-}\r
-\r
-static int stbi__psd_is16(stbi__context *s)\r
-{\r
- int channelCount, depth;\r
- if (stbi__get32be(s) != 0x38425053) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- if (stbi__get16be(s) != 1) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- stbi__skip(s, 6);\r
- channelCount = stbi__get16be(s);\r
- if (channelCount < 0 || channelCount > 16) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- (void) stbi__get32be(s);\r
- (void) stbi__get32be(s);\r
- depth = stbi__get16be(s);\r
- if (depth != 16) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- return 1;\r
-}\r
-#endif\r
-\r
-#ifndef STBI_NO_PIC\r
-static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- int act_comp=0,num_packets=0,chained,dummy;\r
- stbi__pic_packet packets[10];\r
-\r
- if (!x) x = &dummy;\r
- if (!y) y = &dummy;\r
- if (!comp) comp = &dummy;\r
-\r
- if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) {\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
-\r
- stbi__skip(s, 88);\r
-\r
- *x = stbi__get16be(s);\r
- *y = stbi__get16be(s);\r
- if (stbi__at_eof(s)) {\r
- stbi__rewind( s);\r
- return 0;\r
- }\r
- if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
-\r
- stbi__skip(s, 8);\r
-\r
- do {\r
- stbi__pic_packet *packet;\r
-\r
- if (num_packets==sizeof(packets)/sizeof(packets[0]))\r
- return 0;\r
-\r
- packet = &packets[num_packets++];\r
- chained = stbi__get8(s);\r
- packet->size = stbi__get8(s);\r
- packet->type = stbi__get8(s);\r
- packet->channel = stbi__get8(s);\r
- act_comp |= packet->channel;\r
-\r
- if (stbi__at_eof(s)) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- if (packet->size != 8) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- } while (chained);\r
-\r
- *comp = (act_comp & 0x10 ? 4 : 3);\r
-\r
- return 1;\r
-}\r
-#endif\r
-\r
-// *************************************************************************************************\r
-// Portable Gray Map and Portable Pixel Map loader\r
-// by Ken Miller\r
-//\r
-// PGM: http://netpbm.sourceforge.net/doc/pgm.html\r
-// PPM: http://netpbm.sourceforge.net/doc/ppm.html\r
-//\r
-// Known limitations:\r
-// Does not support comments in the header section\r
-// Does not support ASCII image data (formats P2 and P3)\r
-// Does not support 16-bit-per-channel\r
-\r
-#ifndef STBI_NO_PNM\r
-\r
-static int stbi__pnm_test(stbi__context *s)\r
-{\r
- char p, t;\r
- p = (char) stbi__get8(s);\r
- t = (char) stbi__get8(s);\r
- if (p != 'P' || (t != '5' && t != '6')) {\r
- stbi__rewind( s );\r
- return 0;\r
- }\r
- return 1;\r
-}\r
-\r
-static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)\r
-{\r
- stbi_uc *out;\r
- STBI_NOTUSED(ri);\r
-\r
- if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n))\r
- return 0;\r
-\r
- *x = s->img_x;\r
- *y = s->img_y;\r
- if (comp) *comp = s->img_n;\r
-\r
- if (!stbi__mad3sizes_valid(s->img_n, s->img_x, s->img_y, 0))\r
- return stbi__errpuc("too large", "PNM too large");\r
-\r
- out = (stbi_uc *) stbi__malloc_mad3(s->img_n, s->img_x, s->img_y, 0);\r
- if (!out) return stbi__errpuc("outofmem", "Out of memory");\r
- stbi__getn(s, out, s->img_n * s->img_x * s->img_y);\r
-\r
- if (req_comp && req_comp != s->img_n) {\r
- out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y);\r
- if (out == NULL) return out; // stbi__convert_format frees input on failure\r
- }\r
- return out;\r
-}\r
-\r
-static int stbi__pnm_isspace(char c)\r
-{\r
- return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r';\r
-}\r
-\r
-static void stbi__pnm_skip_whitespace(stbi__context *s, char *c)\r
-{\r
- for (;;) {\r
- while (!stbi__at_eof(s) && stbi__pnm_isspace(*c))\r
- *c = (char) stbi__get8(s);\r
-\r
- if (stbi__at_eof(s) || *c != '#')\r
- break;\r
-\r
- while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' )\r
- *c = (char) stbi__get8(s);\r
- }\r
-}\r
-\r
-static int stbi__pnm_isdigit(char c)\r
-{\r
- return c >= '0' && c <= '9';\r
-}\r
-\r
-static int stbi__pnm_getinteger(stbi__context *s, char *c)\r
-{\r
- int value = 0;\r
-\r
- while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) {\r
- value = value*10 + (*c - '0');\r
- *c = (char) stbi__get8(s);\r
- }\r
-\r
- return value;\r
-}\r
-\r
-static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- int maxv, dummy;\r
- char c, p, t;\r
-\r
- if (!x) x = &dummy;\r
- if (!y) y = &dummy;\r
- if (!comp) comp = &dummy;\r
-\r
- stbi__rewind(s);\r
-\r
- // Get identifier\r
- p = (char) stbi__get8(s);\r
- t = (char) stbi__get8(s);\r
- if (p != 'P' || (t != '5' && t != '6')) {\r
- stbi__rewind(s);\r
- return 0;\r
- }\r
-\r
- *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm\r
-\r
- c = (char) stbi__get8(s);\r
- stbi__pnm_skip_whitespace(s, &c);\r
-\r
- *x = stbi__pnm_getinteger(s, &c); // read width\r
- stbi__pnm_skip_whitespace(s, &c);\r
-\r
- *y = stbi__pnm_getinteger(s, &c); // read height\r
- stbi__pnm_skip_whitespace(s, &c);\r
-\r
- maxv = stbi__pnm_getinteger(s, &c); // read max value\r
-\r
- if (maxv > 255)\r
- return stbi__err("max value > 255", "PPM image not 8-bit");\r
- else\r
- return 1;\r
-}\r
-#endif\r
-\r
-static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)\r
-{\r
- #ifndef STBI_NO_JPEG\r
- if (stbi__jpeg_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_PNG\r
- if (stbi__png_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_GIF\r
- if (stbi__gif_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_BMP\r
- if (stbi__bmp_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_PSD\r
- if (stbi__psd_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_PIC\r
- if (stbi__pic_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_PNM\r
- if (stbi__pnm_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_HDR\r
- if (stbi__hdr_info(s, x, y, comp)) return 1;\r
- #endif\r
-\r
- // test tga last because it's a crappy test!\r
- #ifndef STBI_NO_TGA\r
- if (stbi__tga_info(s, x, y, comp))\r
- return 1;\r
- #endif\r
- return stbi__err("unknown image type", "Image not of any known type, or corrupt");\r
-}\r
-\r
-static int stbi__is_16_main(stbi__context *s)\r
-{\r
- #ifndef STBI_NO_PNG\r
- if (stbi__png_is16(s)) return 1;\r
- #endif\r
-\r
- #ifndef STBI_NO_PSD\r
- if (stbi__psd_is16(s)) return 1;\r
- #endif\r
-\r
- return 0;\r
-}\r
-\r
-#ifndef STBI_NO_STDIO\r
-STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp)\r
-{\r
- FILE *f = stbi__fopen(filename, "rb");\r
- int result;\r
- if (!f) return stbi__err("can't fopen", "Unable to open file");\r
- result = stbi_info_from_file(f, x, y, comp);\r
- fclose(f);\r
- return result;\r
-}\r
-\r
-STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp)\r
-{\r
- int r;\r
- stbi__context s;\r
- long pos = ftell(f);\r
- stbi__start_file(&s, f);\r
- r = stbi__info_main(&s,x,y,comp);\r
- fseek(f,pos,SEEK_SET);\r
- return r;\r
-}\r
-\r
-STBIDEF int stbi_is_16_bit(char const *filename)\r
-{\r
- FILE *f = stbi__fopen(filename, "rb");\r
- int result;\r
- if (!f) return stbi__err("can't fopen", "Unable to open file");\r
- result = stbi_is_16_bit_from_file(f);\r
- fclose(f);\r
- return result;\r
-}\r
-\r
-STBIDEF int stbi_is_16_bit_from_file(FILE *f)\r
-{\r
- int r;\r
- stbi__context s;\r
- long pos = ftell(f);\r
- stbi__start_file(&s, f);\r
- r = stbi__is_16_main(&s);\r
- fseek(f,pos,SEEK_SET);\r
- return r;\r
-}\r
-#endif // !STBI_NO_STDIO\r
-\r
-STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp)\r
-{\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__info_main(&s,x,y,comp);\r
-}\r
-\r
-STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp)\r
-{\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);\r
- return stbi__info_main(&s,x,y,comp);\r
-}\r
-\r
-STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len)\r
-{\r
- stbi__context s;\r
- stbi__start_mem(&s,buffer,len);\r
- return stbi__is_16_main(&s);\r
-}\r
-\r
-STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user)\r
-{\r
- stbi__context s;\r
- stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);\r
- return stbi__is_16_main(&s);\r
-}\r
-\r
-#endif // STB_IMAGE_IMPLEMENTATION\r
-\r
-/*\r
- revision history:\r
- 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs \r
- 2.19 (2018-02-11) fix warning\r
- 2.18 (2018-01-30) fix warnings\r
- 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug\r
- 1-bit BMP\r
- *_is_16_bit api\r
- avoid warnings\r
- 2.16 (2017-07-23) all functions have 16-bit variants;\r
- STBI_NO_STDIO works again;\r
- compilation fixes;\r
- fix rounding in unpremultiply;\r
- optimize vertical flip;\r
- disable raw_len validation;\r
- documentation fixes\r
- 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode;\r
- warning fixes; disable run-time SSE detection on gcc;\r
- uniform handling of optional "return" values;\r
- thread-safe initialization of zlib tables\r
- 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs\r
- 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now\r
- 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes\r
- 2.11 (2016-04-02) allocate large structures on the stack\r
- remove white matting for transparent PSD\r
- fix reported channel count for PNG & BMP\r
- re-enable SSE2 in non-gcc 64-bit\r
- support RGB-formatted JPEG\r
- read 16-bit PNGs (only as 8-bit)\r
- 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED\r
- 2.09 (2016-01-16) allow comments in PNM files\r
- 16-bit-per-pixel TGA (not bit-per-component)\r
- info() for TGA could break due to .hdr handling\r
- info() for BMP to shares code instead of sloppy parse\r
- can use STBI_REALLOC_SIZED if allocator doesn't support realloc\r
- code cleanup\r
- 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA\r
- 2.07 (2015-09-13) fix compiler warnings\r
- partial animated GIF support\r
- limited 16-bpc PSD support\r
- #ifdef unused functions\r
- bug with < 92 byte PIC,PNM,HDR,TGA\r
- 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value\r
- 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning\r
- 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit\r
- 2.03 (2015-04-12) extra corruption checking (mmozeiko)\r
- stbi_set_flip_vertically_on_load (nguillemot)\r
- fix NEON support; fix mingw support\r
- 2.02 (2015-01-19) fix incorrect assert, fix warning\r
- 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2\r
- 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG\r
- 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg)\r
- progressive JPEG (stb)\r
- PGM/PPM support (Ken Miller)\r
- STBI_MALLOC,STBI_REALLOC,STBI_FREE\r
- GIF bugfix -- seemingly never worked\r
- STBI_NO_*, STBI_ONLY_*\r
- 1.48 (2014-12-14) fix incorrectly-named assert()\r
- 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb)\r
- optimize PNG (ryg)\r
- fix bug in interlaced PNG with user-specified channel count (stb)\r
- 1.46 (2014-08-26)\r
- fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG\r
- 1.45 (2014-08-16)\r
- fix MSVC-ARM internal compiler error by wrapping malloc\r
- 1.44 (2014-08-07)\r
- various warning fixes from Ronny Chevalier\r
- 1.43 (2014-07-15)\r
- fix MSVC-only compiler problem in code changed in 1.42\r
- 1.42 (2014-07-09)\r
- don't define _CRT_SECURE_NO_WARNINGS (affects user code)\r
- fixes to stbi__cleanup_jpeg path\r
- added STBI_ASSERT to avoid requiring assert.h\r
- 1.41 (2014-06-25)\r
- fix search&replace from 1.36 that messed up comments/error messages\r
- 1.40 (2014-06-22)\r
- fix gcc struct-initialization warning\r
- 1.39 (2014-06-15)\r
- fix to TGA optimization when req_comp != number of components in TGA;\r
- fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite)\r
- add support for BMP version 5 (more ignored fields)\r
- 1.38 (2014-06-06)\r
- suppress MSVC warnings on integer casts truncating values\r
- fix accidental rename of 'skip' field of I/O\r
- 1.37 (2014-06-04)\r
- remove duplicate typedef\r
- 1.36 (2014-06-03)\r
- convert to header file single-file library\r
- if de-iphone isn't set, load iphone images color-swapped instead of returning NULL\r
- 1.35 (2014-05-27)\r
- various warnings\r
- fix broken STBI_SIMD path\r
- fix bug where stbi_load_from_file no longer left file pointer in correct place\r
- fix broken non-easy path for 32-bit BMP (possibly never used)\r
- TGA optimization by Arseny Kapoulkine\r
- 1.34 (unknown)\r
- use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case\r
- 1.33 (2011-07-14)\r
- make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements\r
- 1.32 (2011-07-13)\r
- support for "info" function for all supported filetypes (SpartanJ)\r
- 1.31 (2011-06-20)\r
- a few more leak fixes, bug in PNG handling (SpartanJ)\r
- 1.30 (2011-06-11)\r
- added ability to load files via callbacks to accomidate custom input streams (Ben Wenger)\r
- removed deprecated format-specific test/load functions\r
- removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway\r
- error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha)\r
- fix inefficiency in decoding 32-bit BMP (David Woo)\r
- 1.29 (2010-08-16)\r
- various warning fixes from Aurelien Pocheville\r
- 1.28 (2010-08-01)\r
- fix bug in GIF palette transparency (SpartanJ)\r
- 1.27 (2010-08-01)\r
- cast-to-stbi_uc to fix warnings\r
- 1.26 (2010-07-24)\r
- fix bug in file buffering for PNG reported by SpartanJ\r
- 1.25 (2010-07-17)\r
- refix trans_data warning (Won Chun)\r
- 1.24 (2010-07-12)\r
- perf improvements reading from files on platforms with lock-heavy fgetc()\r
- minor perf improvements for jpeg\r
- deprecated type-specific functions so we'll get feedback if they're needed\r
- attempt to fix trans_data warning (Won Chun)\r
- 1.23 fixed bug in iPhone support\r
- 1.22 (2010-07-10)\r
- removed image *writing* support\r
- stbi_info support from Jetro Lauha\r
- GIF support from Jean-Marc Lienher\r
- iPhone PNG-extensions from James Brown\r
- warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva)\r
- 1.21 fix use of 'stbi_uc' in header (reported by jon blow)\r
- 1.20 added support for Softimage PIC, by Tom Seddon\r
- 1.19 bug in interlaced PNG corruption check (found by ryg)\r
- 1.18 (2008-08-02)\r
- fix a threading bug (local mutable static)\r
- 1.17 support interlaced PNG\r
- 1.16 major bugfix - stbi__convert_format converted one too many pixels\r
- 1.15 initialize some fields for thread safety\r
- 1.14 fix threadsafe conversion bug\r
- header-file-only version (#define STBI_HEADER_FILE_ONLY before including)\r
- 1.13 threadsafe\r
- 1.12 const qualifiers in the API\r
- 1.11 Support installable IDCT, colorspace conversion routines\r
- 1.10 Fixes for 64-bit (don't use "unsigned long")\r
- optimized upsampling by Fabian "ryg" Giesen\r
- 1.09 Fix format-conversion for PSD code (bad global variables!)\r
- 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz\r
- 1.07 attempt to fix C++ warning/errors again\r
- 1.06 attempt to fix C++ warning/errors again\r
- 1.05 fix TGA loading to return correct *comp and use good luminance calc\r
- 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free\r
- 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR\r
- 1.02 support for (subset of) HDR files, float interface for preferred access to them\r
- 1.01 fix bug: possible bug in handling right-side up bmps... not sure\r
- fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all\r
- 1.00 interface to zlib that skips zlib header\r
- 0.99 correct handling of alpha in palette\r
- 0.98 TGA loader by lonesock; dynamically add loaders (untested)\r
- 0.97 jpeg errors on too large a file; also catch another malloc failure\r
- 0.96 fix detection of invalid v value - particleman@mollyrocket forum\r
- 0.95 during header scan, seek to markers in case of padding\r
- 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same\r
- 0.93 handle jpegtran output; verbose errors\r
- 0.92 read 4,8,16,24,32-bit BMP files of several formats\r
- 0.91 output 24-bit Windows 3.0 BMP files\r
- 0.90 fix a few more warnings; bump version number to approach 1.0\r
- 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd\r
- 0.60 fix compiling as c++\r
- 0.59 fix warnings: merge Dave Moore's -Wall fixes\r
- 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian\r
- 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available\r
- 0.56 fix bug: zlib uncompressed mode len vs. nlen\r
- 0.55 fix bug: restart_interval not initialized to 0\r
- 0.54 allow NULL for 'int *comp'\r
- 0.53 fix bug in png 3->4; speedup png decoding\r
- 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments\r
- 0.51 obey req_comp requests, 1-component jpegs return as 1-component,\r
- on 'test' only check type, not whether we support this variant\r
- 0.50 (2006-11-19)\r
- first released version\r
-*/\r
-\r
-\r
-/*\r
-------------------------------------------------------------------------------\r
-This software is available under 2 licenses -- choose whichever you prefer.\r
-------------------------------------------------------------------------------\r
-ALTERNATIVE A - MIT License\r
-Copyright (c) 2017 Sean Barrett\r
-Permission is hereby granted, free of charge, to any person obtaining a copy of\r
-this software and associated documentation files (the "Software"), to deal in\r
-the Software without restriction, including without limitation the rights to\r
-use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies\r
-of the Software, and to permit persons to whom the Software is furnished to do\r
-so, subject to the following conditions:\r
-The above copyright notice and this permission notice shall be included in all\r
-copies or substantial portions of the Software.\r
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\r
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\r
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\r
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\r
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\r
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\r
-SOFTWARE.\r
-------------------------------------------------------------------------------\r
-ALTERNATIVE B - Public Domain (www.unlicense.org)\r
-This is free and unencumbered software released into the public domain.\r
-Anyone is free to copy, modify, publish, use, compile, sell, or distribute this\r
-software, either in source code form or as a compiled binary, for any purpose,\r
-commercial or non-commercial, and by any means.\r
-In jurisdictions that recognize copyright laws, the author or authors of this\r
-software dedicate any and all copyright interest in the software to the public\r
-domain. We make this dedication for the benefit of the public at large and to\r
-the detriment of our heirs and successors. We intend this dedication to be an\r
-overt act of relinquishment in perpetuity of all present and future rights to\r
-this software under copyright law.\r
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\r
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\r
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\r
-AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN\r
-ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION\r
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\r
-------------------------------------------------------------------------------\r
-*/\r
+/* stb_image - v2.23 - public domain image loader - http://nothings.org/stb
+ no warranty implied; use at your own risk
+
+ Do this:
+ #define STB_IMAGE_IMPLEMENTATION
+ before you include this file in *one* C or C++ file to create the implementation.
+
+ // i.e. it should look like this:
+ #include ...
+ #include ...
+ #include ...
+ #define STB_IMAGE_IMPLEMENTATION
+ #include "stb_image.h"
+
+ You can #define STBI_ASSERT(x) before the #include to avoid using assert.h.
+ And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free
+
+
+ QUICK NOTES:
+ Primarily of interest to game developers and other people who can
+ avoid problematic images and only need the trivial interface
+
+ JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib)
+ PNG 1/2/4/8/16-bit-per-channel
+
+ TGA (not sure what subset, if a subset)
+ BMP non-1bpp, non-RLE
+ PSD (composited view only, no extra channels, 8/16 bit-per-channel)
+
+ GIF (*comp always reports as 4-channel)
+ HDR (radiance rgbE format)
+ PIC (Softimage PIC)
+ PNM (PPM and PGM binary only)
+
+ Animated GIF still needs a proper API, but here's one way to do it:
+ http://gist.github.com/urraka/685d9a6340b26b830d49
+
+ - decode from memory or through FILE (define STBI_NO_STDIO to remove code)
+ - decode from arbitrary I/O callbacks
+ - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON)
+
+ Full documentation under "DOCUMENTATION" below.
+
+
+LICENSE
+
+ See end of file for license information.
+
+RECENT REVISION HISTORY:
+
+ 2.23 (2019-08-11) fix clang static analysis warning
+ 2.22 (2019-03-04) gif fixes, fix warnings
+ 2.21 (2019-02-25) fix typo in comment
+ 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs
+ 2.19 (2018-02-11) fix warning
+ 2.18 (2018-01-30) fix warnings
+ 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings
+ 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes
+ 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC
+ 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+ 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes
+ 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+ 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64
+ RGB-format JPEG; remove white matting in PSD;
+ allocate large structures on the stack;
+ correct channel count for PNG & BMP
+ 2.10 (2016-01-22) avoid warning introduced in 2.09
+ 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED
+
+ See end of file for full revision history.
+
+
+ ============================ Contributors =========================
+
+ Image formats Extensions, features
+ Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info)
+ Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info)
+ Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG)
+ Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks)
+ Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG)
+ Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip)
+ Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD)
+ github:urraka (animated gif) Junggon Kim (PNM comments)
+ Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA)
+ socks-the-fox (16-bit PNG)
+ Jeremy Sawicki (handle all ImageNet JPGs)
+ Optimizations & bugfixes Mikhail Morozov (1-bit BMP)
+ Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query)
+ Arseny Kapoulkine
+ John-Mark Allen
+ Carmelo J Fdez-Aguera
+
+ Bug & warning fixes
+ Marc LeBlanc David Woo Guillaume George Martins Mozeiko
+ Christpher Lloyd Jerry Jansson Joseph Thomson Phil Jordan
+ Dave Moore Roy Eltham Hayaki Saito Nathan Reed
+ Won Chun Luke Graham Johan Duparc Nick Verigakis
+ the Horde3D community Thomas Ruf Ronny Chevalier github:rlyeh
+ Janez Zemva John Bartholomew Michal Cichon github:romigrou
+ Jonathan Blow Ken Hamada Tero Hanninen github:svdijk
+ Laurent Gomila Cort Stratton Sergio Gonzalez github:snagar
+ Aruelien Pocheville Thibault Reuille Cass Everitt github:Zelex
+ Ryamond Barbiero Paul Du Bois Engin Manap github:grim210
+ Aldo Culquicondor Philipp Wiesemann Dale Weiler github:sammyhw
+ Oriol Ferrer Mesia Josh Tobin Matthew Gregan github:phprus
+ Julian Raschke Gregory Mullen Baldur Karlsson github:poppolopoppo
+ Christian Floisand Kevin Schmidt JR Smith github:darealshinji
+ Blazej Dariusz Roszkowski github:Michaelangel007
+*/
+
+#ifndef STBI_INCLUDE_STB_IMAGE_H
+#define STBI_INCLUDE_STB_IMAGE_H
+
+// DOCUMENTATION
+//
+// Limitations:
+// - no 12-bit-per-channel JPEG
+// - no JPEGs with arithmetic coding
+// - GIF always returns *comp=4
+//
+// Basic usage (see HDR discussion below for HDR usage):
+// int x,y,n;
+// unsigned char *data = stbi_load(filename, &x, &y, &n, 0);
+// // ... process data if not NULL ...
+// // ... x = width, y = height, n = # 8-bit components per pixel ...
+// // ... replace '0' with '1'..'4' to force that many components per pixel
+// // ... but 'n' will always be the number that it would have been if you said 0
+// stbi_image_free(data)
+//
+// Standard parameters:
+// int *x -- outputs image width in pixels
+// int *y -- outputs image height in pixels
+// int *channels_in_file -- outputs # of image components in image file
+// int desired_channels -- if non-zero, # of image components requested in result
+//
+// The return value from an image loader is an 'unsigned char *' which points
+// to the pixel data, or NULL on an allocation failure or if the image is
+// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels,
+// with each pixel consisting of N interleaved 8-bit components; the first
+// pixel pointed to is top-left-most in the image. There is no padding between
+// image scanlines or between pixels, regardless of format. The number of
+// components N is 'desired_channels' if desired_channels is non-zero, or
+// *channels_in_file otherwise. If desired_channels is non-zero,
+// *channels_in_file has the number of components that _would_ have been
+// output otherwise. E.g. if you set desired_channels to 4, you will always
+// get RGBA output, but you can check *channels_in_file to see if it's trivially
+// opaque because e.g. there were only 3 channels in the source image.
+//
+// An output image with N components has the following components interleaved
+// in this order in each pixel:
+//
+// N=#comp components
+// 1 grey
+// 2 grey, alpha
+// 3 red, green, blue
+// 4 red, green, blue, alpha
+//
+// If image loading fails for any reason, the return value will be NULL,
+// and *x, *y, *channels_in_file will be unchanged. The function
+// stbi_failure_reason() can be queried for an extremely brief, end-user
+// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS
+// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly
+// more user-friendly ones.
+//
+// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.
+//
+// ===========================================================================
+//
+// UNICODE:
+//
+// If compiling for Windows and you wish to use Unicode filenames, compile
+// with
+// #define STBI_WINDOWS_UTF8
+// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert
+// Windows wchar_t filenames to utf8.
+//
+// ===========================================================================
+//
+// Philosophy
+//
+// stb libraries are designed with the following priorities:
+//
+// 1. easy to use
+// 2. easy to maintain
+// 3. good performance
+//
+// Sometimes I let "good performance" creep up in priority over "easy to maintain",
+// and for best performance I may provide less-easy-to-use APIs that give higher
+// performance, in addition to the easy-to-use ones. Nevertheless, it's important
+// to keep in mind that from the standpoint of you, a client of this library,
+// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all.
+//
+// Some secondary priorities arise directly from the first two, some of which
+// provide more explicit reasons why performance can't be emphasized.
+//
+// - Portable ("ease of use")
+// - Small source code footprint ("easy to maintain")
+// - No dependencies ("ease of use")
+//
+// ===========================================================================
+//
+// I/O callbacks
+//
+// I/O callbacks allow you to read from arbitrary sources, like packaged
+// files or some other source. Data read from callbacks are processed
+// through a small internal buffer (currently 128 bytes) to try to reduce
+// overhead.
+//
+// The three functions you must define are "read" (reads some bytes of data),
+// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end).
+//
+// ===========================================================================
+//
+// SIMD support
+//
+// The JPEG decoder will try to automatically use SIMD kernels on x86 when
+// supported by the compiler. For ARM Neon support, you must explicitly
+// request it.
+//
+// (The old do-it-yourself SIMD API is no longer supported in the current
+// code.)
+//
+// On x86, SSE2 will automatically be used when available based on a run-time
+// test; if not, the generic C versions are used as a fall-back. On ARM targets,
+// the typical path is to have separate builds for NEON and non-NEON devices
+// (at least this is true for iOS and Android). Therefore, the NEON support is
+// toggled by a build flag: define STBI_NEON to get NEON loops.
+//
+// If for some reason you do not want to use any of SIMD code, or if
+// you have issues compiling it, you can disable it entirely by
+// defining STBI_NO_SIMD.
+//
+// ===========================================================================
+//
+// HDR image support (disable by defining STBI_NO_HDR)
+//
+// stb_image supports loading HDR images in general, and currently the Radiance
+// .HDR file format specifically. You can still load any file through the existing
+// interface; if you attempt to load an HDR file, it will be automatically remapped
+// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1;
+// both of these constants can be reconfigured through this interface:
+//
+// stbi_hdr_to_ldr_gamma(2.2f);
+// stbi_hdr_to_ldr_scale(1.0f);
+//
+// (note, do not use _inverse_ constants; stbi_image will invert them
+// appropriately).
+//
+// Additionally, there is a new, parallel interface for loading files as
+// (linear) floats to preserve the full dynamic range:
+//
+// float *data = stbi_loadf(filename, &x, &y, &n, 0);
+//
+// If you load LDR images through this interface, those images will
+// be promoted to floating point values, run through the inverse of
+// constants corresponding to the above:
+//
+// stbi_ldr_to_hdr_scale(1.0f);
+// stbi_ldr_to_hdr_gamma(2.2f);
+//
+// Finally, given a filename (or an open file or memory block--see header
+// file for details) containing image data, you can query for the "most
+// appropriate" interface to use (that is, whether the image is HDR or
+// not), using:
+//
+// stbi_is_hdr(char *filename);
+//
+// ===========================================================================
+//
+// iPhone PNG support:
+//
+// By default we convert iphone-formatted PNGs back to RGB, even though
+// they are internally encoded differently. You can disable this conversion
+// by calling stbi_convert_iphone_png_to_rgb(0), in which case
+// you will always just get the native iphone "format" through (which
+// is BGR stored in RGB).
+//
+// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per
+// pixel to remove any premultiplied alpha *only* if the image file explicitly
+// says there's premultiplied data (currently only happens in iPhone images,
+// and only if iPhone convert-to-rgb processing is on).
+//
+// ===========================================================================
+//
+// ADDITIONAL CONFIGURATION
+//
+// - You can suppress implementation of any of the decoders to reduce
+// your code footprint by #defining one or more of the following
+// symbols before creating the implementation.
+//
+// STBI_NO_JPEG
+// STBI_NO_PNG
+// STBI_NO_BMP
+// STBI_NO_PSD
+// STBI_NO_TGA
+// STBI_NO_GIF
+// STBI_NO_HDR
+// STBI_NO_PIC
+// STBI_NO_PNM (.ppm and .pgm)
+//
+// - You can request *only* certain decoders and suppress all other ones
+// (this will be more forward-compatible, as addition of new decoders
+// doesn't require you to disable them explicitly):
+//
+// STBI_ONLY_JPEG
+// STBI_ONLY_PNG
+// STBI_ONLY_BMP
+// STBI_ONLY_PSD
+// STBI_ONLY_TGA
+// STBI_ONLY_GIF
+// STBI_ONLY_HDR
+// STBI_ONLY_PIC
+// STBI_ONLY_PNM (.ppm and .pgm)
+//
+// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still
+// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB
+//
+
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif // STBI_NO_STDIO
+
+#define STBI_VERSION 1
+
+enum
+{
+ STBI_default = 0, // only used for desired_channels
+
+ STBI_grey = 1,
+ STBI_grey_alpha = 2,
+ STBI_rgb = 3,
+ STBI_rgb_alpha = 4
+};
+
+#include <stdlib.h>
+typedef unsigned char stbi_uc;
+typedef unsigned short stbi_us;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef STBIDEF
+#ifdef STB_IMAGE_STATIC
+#define STBIDEF static
+#else
+#define STBIDEF extern
+#endif
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// PRIMARY API - works on images of any type
+//
+
+//
+// load image by filename, open file, or memory buffer
+//
+
+typedef struct
+{
+ int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read
+ void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative
+ int (*eof) (void *user); // returns nonzero if we are at end of file/data
+} stbi_io_callbacks;
+
+////////////////////////////////////
+//
+// 8-bits-per-channel interface
+//
+
+STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+// for stbi_load_from_file, file pointer is left pointing immediately after image
+#endif
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+#endif
+
+#ifdef STBI_WINDOWS_UTF8
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input);
+#endif
+
+////////////////////////////////////
+//
+// 16-bits-per-channel interface
+//
+
+STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+#endif
+
+////////////////////////////////////
+//
+// float-per-channel interface
+//
+#ifndef STBI_NO_LINEAR
+ STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+ STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+ #ifndef STBI_NO_STDIO
+ STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+ STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+ #endif
+#endif
+
+#ifndef STBI_NO_HDR
+ STBIDEF void stbi_hdr_to_ldr_gamma(float gamma);
+ STBIDEF void stbi_hdr_to_ldr_scale(float scale);
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_LINEAR
+ STBIDEF void stbi_ldr_to_hdr_gamma(float gamma);
+ STBIDEF void stbi_ldr_to_hdr_scale(float scale);
+#endif // STBI_NO_LINEAR
+
+// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR
+STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len);
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_is_hdr (char const *filename);
+STBIDEF int stbi_is_hdr_from_file(FILE *f);
+#endif // STBI_NO_STDIO
+
+
+// get a VERY brief reason for failure
+// NOT THREADSAFE
+STBIDEF const char *stbi_failure_reason (void);
+
+// free the loaded image -- this is just free()
+STBIDEF void stbi_image_free (void *retval_from_stbi_load);
+
+// get image dimensions & components without fully decoding
+STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp);
+STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp);
+STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len);
+STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp);
+STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp);
+STBIDEF int stbi_is_16_bit (char const *filename);
+STBIDEF int stbi_is_16_bit_from_file(FILE *f);
+#endif
+
+
+
+// for image formats that explicitly notate that they have premultiplied alpha,
+// we just return the colors as stored in the file. set this flag to force
+// unpremultiplication. results are undefined if the unpremultiply overflow.
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply);
+
+// indicate whether we should process iphone images back to canonical format,
+// or just pass them through "as-is"
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert);
+
+// flip the image vertically, so the first pixel in the output array is the bottom left
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip);
+
+// ZLIB client - used by PNG, available for other purposes
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);
+STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+//
+//
+//// end header file /////////////////////////////////////////////////////
+#endif // STBI_INCLUDE_STB_IMAGE_H
+
+#ifdef STB_IMAGE_IMPLEMENTATION
+
+#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \
+ || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \
+ || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \
+ || defined(STBI_ONLY_ZLIB)
+ #ifndef STBI_ONLY_JPEG
+ #define STBI_NO_JPEG
+ #endif
+ #ifndef STBI_ONLY_PNG
+ #define STBI_NO_PNG
+ #endif
+ #ifndef STBI_ONLY_BMP
+ #define STBI_NO_BMP
+ #endif
+ #ifndef STBI_ONLY_PSD
+ #define STBI_NO_PSD
+ #endif
+ #ifndef STBI_ONLY_TGA
+ #define STBI_NO_TGA
+ #endif
+ #ifndef STBI_ONLY_GIF
+ #define STBI_NO_GIF
+ #endif
+ #ifndef STBI_ONLY_HDR
+ #define STBI_NO_HDR
+ #endif
+ #ifndef STBI_ONLY_PIC
+ #define STBI_NO_PIC
+ #endif
+ #ifndef STBI_ONLY_PNM
+ #define STBI_NO_PNM
+ #endif
+#endif
+
+#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB)
+#define STBI_NO_ZLIB
+#endif
+
+
+#include <stdarg.h>
+#include <stddef.h> // ptrdiff_t on osx
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+#include <math.h> // ldexp, pow
+#endif
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif
+
+#ifndef STBI_ASSERT
+#include <assert.h>
+#define STBI_ASSERT(x) assert(x)
+#endif
+
+#ifdef __cplusplus
+#define STBI_EXTERN extern "C"
+#else
+#define STBI_EXTERN extern
+#endif
+
+
+#ifndef _MSC_VER
+ #ifdef __cplusplus
+ #define stbi_inline inline
+ #else
+ #define stbi_inline
+ #endif
+#else
+ #define stbi_inline __forceinline
+#endif
+
+
+#ifdef _MSC_VER
+typedef unsigned short stbi__uint16;
+typedef signed short stbi__int16;
+typedef unsigned int stbi__uint32;
+typedef signed int stbi__int32;
+#else
+#include <stdint.h>
+typedef uint16_t stbi__uint16;
+typedef int16_t stbi__int16;
+typedef uint32_t stbi__uint32;
+typedef int32_t stbi__int32;
+#endif
+
+// should produce compiler error if size is wrong
+typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1];
+
+#ifdef _MSC_VER
+#define STBI_NOTUSED(v) (void)(v)
+#else
+#define STBI_NOTUSED(v) (void)sizeof(v)
+#endif
+
+#ifdef _MSC_VER
+#define STBI_HAS_LROTL
+#endif
+
+#ifdef STBI_HAS_LROTL
+ #define stbi_lrot(x,y) _lrotl(x,y)
+#else
+ #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y))))
+#endif
+
+#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED))
+// ok
+#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED)
+// ok
+#else
+#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)."
+#endif
+
+#ifndef STBI_MALLOC
+#define STBI_MALLOC(sz) malloc(sz)
+#define STBI_REALLOC(p,newsz) realloc(p,newsz)
+#define STBI_FREE(p) free(p)
+#endif
+
+#ifndef STBI_REALLOC_SIZED
+#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz)
+#endif
+
+// x86/x64 detection
+#if defined(__x86_64__) || defined(_M_X64)
+#define STBI__X64_TARGET
+#elif defined(__i386) || defined(_M_IX86)
+#define STBI__X86_TARGET
+#endif
+
+#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD)
+// gcc doesn't support sse2 intrinsics unless you compile with -msse2,
+// which in turn means it gets to use SSE2 everywhere. This is unfortunate,
+// but previous attempts to provide the SSE2 functions with runtime
+// detection caused numerous issues. The way architecture extensions are
+// exposed in GCC/Clang is, sadly, not really suited for one-file libs.
+// New behavior: if compiled with -msse2, we use SSE2 without any
+// detection; if not, we don't use it at all.
+#define STBI_NO_SIMD
+#endif
+
+#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD)
+// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET
+//
+// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the
+// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant.
+// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not
+// simultaneously enabling "-mstackrealign".
+//
+// See https://github.com/nothings/stb/issues/81 for more information.
+//
+// So default to no SSE2 on 32-bit MinGW. If you've read this far and added
+// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2.
+#define STBI_NO_SIMD
+#endif
+
+#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET))
+#define STBI_SSE2
+#include <emmintrin.h>
+
+#ifdef _MSC_VER
+
+#if _MSC_VER >= 1400 // not VC6
+#include <intrin.h> // __cpuid
+static int stbi__cpuid3(void)
+{
+ int info[4];
+ __cpuid(info,1);
+ return info[3];
+}
+#else
+static int stbi__cpuid3(void)
+{
+ int res;
+ __asm {
+ mov eax,1
+ cpuid
+ mov res,edx
+ }
+ return res;
+}
+#endif
+
+#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+ int info3 = stbi__cpuid3();
+ return ((info3 >> 26) & 1) != 0;
+}
+#endif
+
+#else // assume GCC-style if not VC++
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+ // If we're even attempting to compile this on GCC/Clang, that means
+ // -msse2 is on, which means the compiler is allowed to use SSE2
+ // instructions at will, and so are we.
+ return 1;
+}
+#endif
+
+#endif
+#endif
+
+// ARM NEON
+#if defined(STBI_NO_SIMD) && defined(STBI_NEON)
+#undef STBI_NEON
+#endif
+
+#ifdef STBI_NEON
+#include <arm_neon.h>
+// assume GCC or Clang on ARM targets
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+#endif
+
+#ifndef STBI_SIMD_ALIGN
+#define STBI_SIMD_ALIGN(type, name) type name
+#endif
+
+///////////////////////////////////////////////
+//
+// stbi__context struct and start_xxx functions
+
+// stbi__context structure is our basic context used by all images, so it
+// contains all the IO context, plus some basic image information
+typedef struct
+{
+ stbi__uint32 img_x, img_y;
+ int img_n, img_out_n;
+
+ stbi_io_callbacks io;
+ void *io_user_data;
+
+ int read_from_callbacks;
+ int buflen;
+ stbi_uc buffer_start[128];
+
+ stbi_uc *img_buffer, *img_buffer_end;
+ stbi_uc *img_buffer_original, *img_buffer_original_end;
+} stbi__context;
+
+
+static void stbi__refill_buffer(stbi__context *s);
+
+// initialize a memory-decode context
+static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len)
+{
+ s->io.read = NULL;
+ s->read_from_callbacks = 0;
+ s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer;
+ s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len;
+}
+
+// initialize a callback-based context
+static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)
+{
+ s->io = *c;
+ s->io_user_data = user;
+ s->buflen = sizeof(s->buffer_start);
+ s->read_from_callbacks = 1;
+ s->img_buffer_original = s->buffer_start;
+ stbi__refill_buffer(s);
+ s->img_buffer_original_end = s->img_buffer_end;
+}
+
+#ifndef STBI_NO_STDIO
+
+static int stbi__stdio_read(void *user, char *data, int size)
+{
+ return (int) fread(data,1,size,(FILE*) user);
+}
+
+static void stbi__stdio_skip(void *user, int n)
+{
+ fseek((FILE*) user, n, SEEK_CUR);
+}
+
+static int stbi__stdio_eof(void *user)
+{
+ return feof((FILE*) user);
+}
+
+static stbi_io_callbacks stbi__stdio_callbacks =
+{
+ stbi__stdio_read,
+ stbi__stdio_skip,
+ stbi__stdio_eof,
+};
+
+static void stbi__start_file(stbi__context *s, FILE *f)
+{
+ stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);
+}
+
+//static void stop_file(stbi__context *s) { }
+
+#endif // !STBI_NO_STDIO
+
+static void stbi__rewind(stbi__context *s)
+{
+ // conceptually rewind SHOULD rewind to the beginning of the stream,
+ // but we just rewind to the beginning of the initial buffer, because
+ // we only use it after doing 'test', which only ever looks at at most 92 bytes
+ s->img_buffer = s->img_buffer_original;
+ s->img_buffer_end = s->img_buffer_original_end;
+}
+
+enum
+{
+ STBI_ORDER_RGB,
+ STBI_ORDER_BGR
+};
+
+typedef struct
+{
+ int bits_per_channel;
+ int num_channels;
+ int channel_order;
+} stbi__result_info;
+
+#ifndef STBI_NO_JPEG
+static int stbi__jpeg_test(stbi__context *s);
+static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNG
+static int stbi__png_test(stbi__context *s);
+static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp);
+static int stbi__png_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_test(stbi__context *s);
+static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_TGA
+static int stbi__tga_test(stbi__context *s);
+static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_test(stbi__context *s);
+static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc);
+static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp);
+static int stbi__psd_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_HDR
+static int stbi__hdr_test(stbi__context *s);
+static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_test(stbi__context *s);
+static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_GIF
+static int stbi__gif_test(stbi__context *s);
+static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNM
+static int stbi__pnm_test(stbi__context *s);
+static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+// this is not threadsafe
+static const char *stbi__g_failure_reason;
+
+STBIDEF const char *stbi_failure_reason(void)
+{
+ return stbi__g_failure_reason;
+}
+
+static int stbi__err(const char *str)
+{
+ stbi__g_failure_reason = str;
+ return 0;
+}
+
+static void *stbi__malloc(size_t size)
+{
+ return STBI_MALLOC(size);
+}
+
+// stb_image uses ints pervasively, including for offset calculations.
+// therefore the largest decoded image size we can support with the
+// current code, even on 64-bit targets, is INT_MAX. this is not a
+// significant limitation for the intended use case.
+//
+// we do, however, need to make sure our size calculations don't
+// overflow. hence a few helper functions for size calculations that
+// multiply integers together, making sure that they're non-negative
+// and no overflow occurs.
+
+// return 1 if the sum is valid, 0 on overflow.
+// negative terms are considered invalid.
+static int stbi__addsizes_valid(int a, int b)
+{
+ if (b < 0) return 0;
+ // now 0 <= b <= INT_MAX, hence also
+ // 0 <= INT_MAX - b <= INTMAX.
+ // And "a + b <= INT_MAX" (which might overflow) is the
+ // same as a <= INT_MAX - b (no overflow)
+ return a <= INT_MAX - b;
+}
+
+// returns 1 if the product is valid, 0 on overflow.
+// negative factors are considered invalid.
+static int stbi__mul2sizes_valid(int a, int b)
+{
+ if (a < 0 || b < 0) return 0;
+ if (b == 0) return 1; // mul-by-0 is always safe
+ // portable way to check for no overflows in a*b
+ return a <= INT_MAX/b;
+}
+
+// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad2sizes_valid(int a, int b, int add)
+{
+ return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add);
+}
+
+// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad3sizes_valid(int a, int b, int c, int add)
+{
+ return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+ stbi__addsizes_valid(a*b*c, add);
+}
+
+// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add)
+{
+ return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+ stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add);
+}
+#endif
+
+// mallocs with size overflow checking
+static void *stbi__malloc_mad2(int a, int b, int add)
+{
+ if (!stbi__mad2sizes_valid(a, b, add)) return NULL;
+ return stbi__malloc(a*b + add);
+}
+
+static void *stbi__malloc_mad3(int a, int b, int c, int add)
+{
+ if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL;
+ return stbi__malloc(a*b*c + add);
+}
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+static void *stbi__malloc_mad4(int a, int b, int c, int d, int add)
+{
+ if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL;
+ return stbi__malloc(a*b*c*d + add);
+}
+#endif
+
+// stbi__err - error
+// stbi__errpf - error returning pointer to float
+// stbi__errpuc - error returning pointer to unsigned char
+
+#ifdef STBI_NO_FAILURE_STRINGS
+ #define stbi__err(x,y) 0
+#elif defined(STBI_FAILURE_USERMSG)
+ #define stbi__err(x,y) stbi__err(y)
+#else
+ #define stbi__err(x,y) stbi__err(x)
+#endif
+
+#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL))
+#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))
+
+STBIDEF void stbi_image_free(void *retval_from_stbi_load)
+{
+ STBI_FREE(retval_from_stbi_load);
+}
+
+#ifndef STBI_NO_LINEAR
+static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp);
+#endif
+
+#ifndef STBI_NO_HDR
+static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp);
+#endif
+
+static int stbi__vertically_flip_on_load = 0;
+
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip)
+{
+ stbi__vertically_flip_on_load = flag_true_if_should_flip;
+}
+
+static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+ memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields
+ ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed
+ ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order
+ ri->num_channels = 0;
+
+ #ifndef STBI_NO_JPEG
+ if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri);
+ #endif
+ #ifndef STBI_NO_PNG
+ if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri);
+ #endif
+ #ifndef STBI_NO_BMP
+ if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri);
+ #endif
+ #ifndef STBI_NO_GIF
+ if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri);
+ #endif
+ #ifndef STBI_NO_PSD
+ if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc);
+ #endif
+ #ifndef STBI_NO_PIC
+ if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri);
+ #endif
+ #ifndef STBI_NO_PNM
+ if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri);
+ #endif
+
+ #ifndef STBI_NO_HDR
+ if (stbi__hdr_test(s)) {
+ float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri);
+ return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp);
+ }
+ #endif
+
+ #ifndef STBI_NO_TGA
+ // test tga last because it's a crappy test!
+ if (stbi__tga_test(s))
+ return stbi__tga_load(s,x,y,comp,req_comp, ri);
+ #endif
+
+ return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels)
+{
+ int i;
+ int img_len = w * h * channels;
+ stbi_uc *reduced;
+
+ reduced = (stbi_uc *) stbi__malloc(img_len);
+ if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory");
+
+ for (i = 0; i < img_len; ++i)
+ reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling
+
+ STBI_FREE(orig);
+ return reduced;
+}
+
+static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels)
+{
+ int i;
+ int img_len = w * h * channels;
+ stbi__uint16 *enlarged;
+
+ enlarged = (stbi__uint16 *) stbi__malloc(img_len*2);
+ if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+
+ for (i = 0; i < img_len; ++i)
+ enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff
+
+ STBI_FREE(orig);
+ return enlarged;
+}
+
+static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel)
+{
+ int row;
+ size_t bytes_per_row = (size_t)w * bytes_per_pixel;
+ stbi_uc temp[2048];
+ stbi_uc *bytes = (stbi_uc *)image;
+
+ for (row = 0; row < (h>>1); row++) {
+ stbi_uc *row0 = bytes + row*bytes_per_row;
+ stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row;
+ // swap row0 with row1
+ size_t bytes_left = bytes_per_row;
+ while (bytes_left) {
+ size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);
+ memcpy(temp, row0, bytes_copy);
+ memcpy(row0, row1, bytes_copy);
+ memcpy(row1, temp, bytes_copy);
+ row0 += bytes_copy;
+ row1 += bytes_copy;
+ bytes_left -= bytes_copy;
+ }
+ }
+}
+
+#ifndef STBI_NO_GIF
+static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel)
+{
+ int slice;
+ int slice_size = w * h * bytes_per_pixel;
+
+ stbi_uc *bytes = (stbi_uc *)image;
+ for (slice = 0; slice < z; ++slice) {
+ stbi__vertical_flip(bytes, w, h, bytes_per_pixel);
+ bytes += slice_size;
+ }
+}
+#endif
+
+static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__result_info ri;
+ void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8);
+
+ if (result == NULL)
+ return NULL;
+
+ if (ri.bits_per_channel != 8) {
+ STBI_ASSERT(ri.bits_per_channel == 16);
+ result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+ ri.bits_per_channel = 8;
+ }
+
+ // @TODO: move stbi__convert_format to here
+
+ if (stbi__vertically_flip_on_load) {
+ int channels = req_comp ? req_comp : *comp;
+ stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc));
+ }
+
+ return (unsigned char *) result;
+}
+
+static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__result_info ri;
+ void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16);
+
+ if (result == NULL)
+ return NULL;
+
+ if (ri.bits_per_channel != 16) {
+ STBI_ASSERT(ri.bits_per_channel == 8);
+ result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+ ri.bits_per_channel = 16;
+ }
+
+ // @TODO: move stbi__convert_format16 to here
+ // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision
+
+ if (stbi__vertically_flip_on_load) {
+ int channels = req_comp ? req_comp : *comp;
+ stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16));
+ }
+
+ return (stbi__uint16 *) result;
+}
+
+#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR)
+static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)
+{
+ if (stbi__vertically_flip_on_load && result != NULL) {
+ int channels = req_comp ? req_comp : *comp;
+ stbi__vertical_flip(result, *x, *y, channels * sizeof(float));
+ }
+}
+#endif
+
+#ifndef STBI_NO_STDIO
+
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide);
+STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default);
+#endif
+
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input)
+{
+ return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL);
+}
+#endif
+
+static FILE *stbi__fopen(char const *filename, char const *mode)
+{
+ FILE *f;
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+ wchar_t wMode[64];
+ wchar_t wFilename[1024];
+ if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)))
+ return 0;
+
+ if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)))
+ return 0;
+
+#if _MSC_VER >= 1400
+ if (0 != _wfopen_s(&f, wFilename, wMode))
+ f = 0;
+#else
+ f = _wfopen(wFilename, wMode);
+#endif
+
+#elif defined(_MSC_VER) && _MSC_VER >= 1400
+ if (0 != fopen_s(&f, filename, mode))
+ f=0;
+#else
+ f = fopen(filename, mode);
+#endif
+ return f;
+}
+
+
+STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+ FILE *f = stbi__fopen(filename, "rb");
+ unsigned char *result;
+ if (!f) return stbi__errpuc("can't fopen", "Unable to open file");
+ result = stbi_load_from_file(f,x,y,comp,req_comp);
+ fclose(f);
+ return result;
+}
+
+STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+ unsigned char *result;
+ stbi__context s;
+ stbi__start_file(&s,f);
+ result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+ if (result) {
+ // need to 'unget' all the characters in the IO buffer
+ fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+ }
+ return result;
+}
+
+STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__uint16 *result;
+ stbi__context s;
+ stbi__start_file(&s,f);
+ result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp);
+ if (result) {
+ // need to 'unget' all the characters in the IO buffer
+ fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+ }
+ return result;
+}
+
+STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+ FILE *f = stbi__fopen(filename, "rb");
+ stbi__uint16 *result;
+ if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file");
+ result = stbi_load_from_file_16(f,x,y,comp,req_comp);
+ fclose(f);
+ return result;
+}
+
+
+#endif //!STBI_NO_STDIO
+
+STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user);
+ return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+ return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+ unsigned char *result;
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+
+ result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp);
+ if (stbi__vertically_flip_on_load) {
+ stbi__vertical_flip_slices( result, *x, *y, *z, *comp );
+ }
+
+ return result;
+}
+#endif
+
+#ifndef STBI_NO_LINEAR
+static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+ unsigned char *data;
+ #ifndef STBI_NO_HDR
+ if (stbi__hdr_test(s)) {
+ stbi__result_info ri;
+ float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri);
+ if (hdr_data)
+ stbi__float_postprocess(hdr_data,x,y,comp,req_comp);
+ return hdr_data;
+ }
+ #endif
+ data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp);
+ if (data)
+ return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp);
+ return stbi__errpf("unknown image type", "Image not of any known type, or corrupt");
+}
+
+STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+ return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+ float *result;
+ FILE *f = stbi__fopen(filename, "rb");
+ if (!f) return stbi__errpf("can't fopen", "Unable to open file");
+ result = stbi_loadf_from_file(f,x,y,comp,req_comp);
+ fclose(f);
+ return result;
+}
+
+STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+ stbi__context s;
+ stbi__start_file(&s,f);
+ return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+#endif // !STBI_NO_STDIO
+
+#endif // !STBI_NO_LINEAR
+
+// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is
+// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always
+// reports false!
+
+STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len)
+{
+ #ifndef STBI_NO_HDR
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__hdr_test(&s);
+ #else
+ STBI_NOTUSED(buffer);
+ STBI_NOTUSED(len);
+ return 0;
+ #endif
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_is_hdr (char const *filename)
+{
+ FILE *f = stbi__fopen(filename, "rb");
+ int result=0;
+ if (f) {
+ result = stbi_is_hdr_from_file(f);
+ fclose(f);
+ }
+ return result;
+}
+
+STBIDEF int stbi_is_hdr_from_file(FILE *f)
+{
+ #ifndef STBI_NO_HDR
+ long pos = ftell(f);
+ int res;
+ stbi__context s;
+ stbi__start_file(&s,f);
+ res = stbi__hdr_test(&s);
+ fseek(f, pos, SEEK_SET);
+ return res;
+ #else
+ STBI_NOTUSED(f);
+ return 0;
+ #endif
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user)
+{
+ #ifndef STBI_NO_HDR
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+ return stbi__hdr_test(&s);
+ #else
+ STBI_NOTUSED(clbk);
+ STBI_NOTUSED(user);
+ return 0;
+ #endif
+}
+
+#ifndef STBI_NO_LINEAR
+static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f;
+
+STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; }
+STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; }
+#endif
+
+static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f;
+
+STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; }
+STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; }
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Common code used by all image loaders
+//
+
+enum
+{
+ STBI__SCAN_load=0,
+ STBI__SCAN_type,
+ STBI__SCAN_header
+};
+
+static void stbi__refill_buffer(stbi__context *s)
+{
+ int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);
+ if (n == 0) {
+ // at end of file, treat same as if from memory, but need to handle case
+ // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file
+ s->read_from_callbacks = 0;
+ s->img_buffer = s->buffer_start;
+ s->img_buffer_end = s->buffer_start+1;
+ *s->img_buffer = 0;
+ } else {
+ s->img_buffer = s->buffer_start;
+ s->img_buffer_end = s->buffer_start + n;
+ }
+}
+
+stbi_inline static stbi_uc stbi__get8(stbi__context *s)
+{
+ if (s->img_buffer < s->img_buffer_end)
+ return *s->img_buffer++;
+ if (s->read_from_callbacks) {
+ stbi__refill_buffer(s);
+ return *s->img_buffer++;
+ }
+ return 0;
+}
+
+stbi_inline static int stbi__at_eof(stbi__context *s)
+{
+ if (s->io.read) {
+ if (!(s->io.eof)(s->io_user_data)) return 0;
+ // if feof() is true, check if buffer = end
+ // special case: we've only got the special 0 character at the end
+ if (s->read_from_callbacks == 0) return 1;
+ }
+
+ return s->img_buffer >= s->img_buffer_end;
+}
+
+static void stbi__skip(stbi__context *s, int n)
+{
+ if (n < 0) {
+ s->img_buffer = s->img_buffer_end;
+ return;
+ }
+ if (s->io.read) {
+ int blen = (int) (s->img_buffer_end - s->img_buffer);
+ if (blen < n) {
+ s->img_buffer = s->img_buffer_end;
+ (s->io.skip)(s->io_user_data, n - blen);
+ return;
+ }
+ }
+ s->img_buffer += n;
+}
+
+static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n)
+{
+ if (s->io.read) {
+ int blen = (int) (s->img_buffer_end - s->img_buffer);
+ if (blen < n) {
+ int res, count;
+
+ memcpy(buffer, s->img_buffer, blen);
+
+ count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);
+ res = (count == (n-blen));
+ s->img_buffer = s->img_buffer_end;
+ return res;
+ }
+ }
+
+ if (s->img_buffer+n <= s->img_buffer_end) {
+ memcpy(buffer, s->img_buffer, n);
+ s->img_buffer += n;
+ return 1;
+ } else
+ return 0;
+}
+
+static int stbi__get16be(stbi__context *s)
+{
+ int z = stbi__get8(s);
+ return (z << 8) + stbi__get8(s);
+}
+
+static stbi__uint32 stbi__get32be(stbi__context *s)
+{
+ stbi__uint32 z = stbi__get16be(s);
+ return (z << 16) + stbi__get16be(s);
+}
+
+#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF)
+// nothing
+#else
+static int stbi__get16le(stbi__context *s)
+{
+ int z = stbi__get8(s);
+ return z + (stbi__get8(s) << 8);
+}
+#endif
+
+#ifndef STBI_NO_BMP
+static stbi__uint32 stbi__get32le(stbi__context *s)
+{
+ stbi__uint32 z = stbi__get16le(s);
+ return z + (stbi__get16le(s) << 16);
+}
+#endif
+
+#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// generic converter from built-in img_n to req_comp
+// individual types do this automatically as much as possible (e.g. jpeg
+// does all cases internally since it needs to colorspace convert anyway,
+// and it never has alpha, so very few cases ). png can automatically
+// interleave an alpha=255 channel, but falls back to this for other cases
+//
+// assume data buffer is malloced, so malloc a new one and free that one
+// only failure mode is malloc failing
+
+static stbi_uc stbi__compute_y(int r, int g, int b)
+{
+ return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8);
+}
+
+static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+ int i,j;
+ unsigned char *good;
+
+ if (req_comp == img_n) return data;
+ STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+ good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0);
+ if (good == NULL) {
+ STBI_FREE(data);
+ return stbi__errpuc("outofmem", "Out of memory");
+ }
+
+ for (j=0; j < (int) y; ++j) {
+ unsigned char *src = data + j * x * img_n ;
+ unsigned char *dest = good + j * x * req_comp;
+
+ #define STBI__COMBO(a,b) ((a)*8+(b))
+ #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+ // convert source image with img_n components to one with req_comp components;
+ // avoid switch per pixel, so use switch per scanline and massive macros
+ switch (STBI__COMBO(img_n, req_comp)) {
+ STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break;
+ STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
+ STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break;
+ STBI__CASE(2,1) { dest[0]=src[0]; } break;
+ STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
+ STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;
+ STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break;
+ STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;
+ STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break;
+ STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;
+ STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+ STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;
+ default: STBI_ASSERT(0);
+ }
+ #undef STBI__CASE
+ }
+
+ STBI_FREE(data);
+ return good;
+}
+
+static stbi__uint16 stbi__compute_y_16(int r, int g, int b)
+{
+ return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8);
+}
+
+static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+ int i,j;
+ stbi__uint16 *good;
+
+ if (req_comp == img_n) return data;
+ STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+ good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2);
+ if (good == NULL) {
+ STBI_FREE(data);
+ return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+ }
+
+ for (j=0; j < (int) y; ++j) {
+ stbi__uint16 *src = data + j * x * img_n ;
+ stbi__uint16 *dest = good + j * x * req_comp;
+
+ #define STBI__COMBO(a,b) ((a)*8+(b))
+ #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+ // convert source image with img_n components to one with req_comp components;
+ // avoid switch per pixel, so use switch per scanline and massive macros
+ switch (STBI__COMBO(img_n, req_comp)) {
+ STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break;
+ STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
+ STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break;
+ STBI__CASE(2,1) { dest[0]=src[0]; } break;
+ STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
+ STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;
+ STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break;
+ STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;
+ STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break;
+ STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;
+ STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+ STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;
+ default: STBI_ASSERT(0);
+ }
+ #undef STBI__CASE
+ }
+
+ STBI_FREE(data);
+ return good;
+}
+
+#ifndef STBI_NO_LINEAR
+static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp)
+{
+ int i,k,n;
+ float *output;
+ if (!data) return NULL;
+ output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0);
+ if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); }
+ // compute number of non-alpha components
+ if (comp & 1) n = comp; else n = comp-1;
+ for (i=0; i < x*y; ++i) {
+ for (k=0; k < n; ++k) {
+ output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale);
+ }
+ }
+ if (n < comp) {
+ for (i=0; i < x*y; ++i) {
+ output[i*comp + n] = data[i*comp + n]/255.0f;
+ }
+ }
+ STBI_FREE(data);
+ return output;
+}
+#endif
+
+#ifndef STBI_NO_HDR
+#define stbi__float2int(x) ((int) (x))
+static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp)
+{
+ int i,k,n;
+ stbi_uc *output;
+ if (!data) return NULL;
+ output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0);
+ if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); }
+ // compute number of non-alpha components
+ if (comp & 1) n = comp; else n = comp-1;
+ for (i=0; i < x*y; ++i) {
+ for (k=0; k < n; ++k) {
+ float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f;
+ if (z < 0) z = 0;
+ if (z > 255) z = 255;
+ output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+ }
+ if (k < comp) {
+ float z = data[i*comp+k] * 255 + 0.5f;
+ if (z < 0) z = 0;
+ if (z > 255) z = 255;
+ output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+ }
+ }
+ STBI_FREE(data);
+ return output;
+}
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// "baseline" JPEG/JFIF decoder
+//
+// simple implementation
+// - doesn't support delayed output of y-dimension
+// - simple interface (only one output format: 8-bit interleaved RGB)
+// - doesn't try to recover corrupt jpegs
+// - doesn't allow partial loading, loading multiple at once
+// - still fast on x86 (copying globals into locals doesn't help x86)
+// - allocates lots of intermediate memory (full size of all components)
+// - non-interleaved case requires this anyway
+// - allows good upsampling (see next)
+// high-quality
+// - upsampled channels are bilinearly interpolated, even across blocks
+// - quality integer IDCT derived from IJG's 'slow'
+// performance
+// - fast huffman; reasonable integer IDCT
+// - some SIMD kernels for common paths on targets with SSE2/NEON
+// - uses a lot of intermediate memory, could cache poorly
+
+#ifndef STBI_NO_JPEG
+
+// huffman decoding acceleration
+#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache
+
+typedef struct
+{
+ stbi_uc fast[1 << FAST_BITS];
+ // weirdly, repacking this into AoS is a 10% speed loss, instead of a win
+ stbi__uint16 code[256];
+ stbi_uc values[256];
+ stbi_uc size[257];
+ unsigned int maxcode[18];
+ int delta[17]; // old 'firstsymbol' - old 'firstcode'
+} stbi__huffman;
+
+typedef struct
+{
+ stbi__context *s;
+ stbi__huffman huff_dc[4];
+ stbi__huffman huff_ac[4];
+ stbi__uint16 dequant[4][64];
+ stbi__int16 fast_ac[4][1 << FAST_BITS];
+
+// sizes for components, interleaved MCUs
+ int img_h_max, img_v_max;
+ int img_mcu_x, img_mcu_y;
+ int img_mcu_w, img_mcu_h;
+
+// definition of jpeg image component
+ struct
+ {
+ int id;
+ int h,v;
+ int tq;
+ int hd,ha;
+ int dc_pred;
+
+ int x,y,w2,h2;
+ stbi_uc *data;
+ void *raw_data, *raw_coeff;
+ stbi_uc *linebuf;
+ short *coeff; // progressive only
+ int coeff_w, coeff_h; // number of 8x8 coefficient blocks
+ } img_comp[4];
+
+ stbi__uint32 code_buffer; // jpeg entropy-coded buffer
+ int code_bits; // number of valid bits
+ unsigned char marker; // marker seen while filling entropy buffer
+ int nomore; // flag if we saw a marker so must stop
+
+ int progressive;
+ int spec_start;
+ int spec_end;
+ int succ_high;
+ int succ_low;
+ int eob_run;
+ int jfif;
+ int app14_color_transform; // Adobe APP14 tag
+ int rgb;
+
+ int scan_n, order[4];
+ int restart_interval, todo;
+
+// kernels
+ void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]);
+ void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step);
+ stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs);
+} stbi__jpeg;
+
+static int stbi__build_huffman(stbi__huffman *h, int *count)
+{
+ int i,j,k=0;
+ unsigned int code;
+ // build size list for each symbol (from JPEG spec)
+ for (i=0; i < 16; ++i)
+ for (j=0; j < count[i]; ++j)
+ h->size[k++] = (stbi_uc) (i+1);
+ h->size[k] = 0;
+
+ // compute actual symbols (from jpeg spec)
+ code = 0;
+ k = 0;
+ for(j=1; j <= 16; ++j) {
+ // compute delta to add to code to compute symbol id
+ h->delta[j] = k - code;
+ if (h->size[k] == j) {
+ while (h->size[k] == j)
+ h->code[k++] = (stbi__uint16) (code++);
+ if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG");
+ }
+ // compute largest code + 1 for this size, preshifted as needed later
+ h->maxcode[j] = code << (16-j);
+ code <<= 1;
+ }
+ h->maxcode[j] = 0xffffffff;
+
+ // build non-spec acceleration table; 255 is flag for not-accelerated
+ memset(h->fast, 255, 1 << FAST_BITS);
+ for (i=0; i < k; ++i) {
+ int s = h->size[i];
+ if (s <= FAST_BITS) {
+ int c = h->code[i] << (FAST_BITS-s);
+ int m = 1 << (FAST_BITS-s);
+ for (j=0; j < m; ++j) {
+ h->fast[c+j] = (stbi_uc) i;
+ }
+ }
+ }
+ return 1;
+}
+
+// build a table that decodes both magnitude and value of small ACs in
+// one go.
+static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h)
+{
+ int i;
+ for (i=0; i < (1 << FAST_BITS); ++i) {
+ stbi_uc fast = h->fast[i];
+ fast_ac[i] = 0;
+ if (fast < 255) {
+ int rs = h->values[fast];
+ int run = (rs >> 4) & 15;
+ int magbits = rs & 15;
+ int len = h->size[fast];
+
+ if (magbits && len + magbits <= FAST_BITS) {
+ // magnitude code followed by receive_extend code
+ int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits);
+ int m = 1 << (magbits - 1);
+ if (k < m) k += (~0U << magbits) + 1;
+ // if the result is small enough, we can fit it in fast_ac table
+ if (k >= -128 && k <= 127)
+ fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits));
+ }
+ }
+ }
+}
+
+static void stbi__grow_buffer_unsafe(stbi__jpeg *j)
+{
+ do {
+ unsigned int b = j->nomore ? 0 : stbi__get8(j->s);
+ if (b == 0xff) {
+ int c = stbi__get8(j->s);
+ while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes
+ if (c != 0) {
+ j->marker = (unsigned char) c;
+ j->nomore = 1;
+ return;
+ }
+ }
+ j->code_buffer |= b << (24 - j->code_bits);
+ j->code_bits += 8;
+ } while (j->code_bits <= 24);
+}
+
+// (1 << n) - 1
+static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535};
+
+// decode a jpeg huffman value from the bitstream
+stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h)
+{
+ unsigned int temp;
+ int c,k;
+
+ if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+ // look at the top FAST_BITS and determine what symbol ID it is,
+ // if the code is <= FAST_BITS
+ c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+ k = h->fast[c];
+ if (k < 255) {
+ int s = h->size[k];
+ if (s > j->code_bits)
+ return -1;
+ j->code_buffer <<= s;
+ j->code_bits -= s;
+ return h->values[k];
+ }
+
+ // naive test is to shift the code_buffer down so k bits are
+ // valid, then test against maxcode. To speed this up, we've
+ // preshifted maxcode left so that it has (16-k) 0s at the
+ // end; in other words, regardless of the number of bits, it
+ // wants to be compared against something shifted to have 16;
+ // that way we don't need to shift inside the loop.
+ temp = j->code_buffer >> 16;
+ for (k=FAST_BITS+1 ; ; ++k)
+ if (temp < h->maxcode[k])
+ break;
+ if (k == 17) {
+ // error! code not found
+ j->code_bits -= 16;
+ return -1;
+ }
+
+ if (k > j->code_bits)
+ return -1;
+
+ // convert the huffman code to the symbol id
+ c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k];
+ STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]);
+
+ // convert the id to a symbol
+ j->code_bits -= k;
+ j->code_buffer <<= k;
+ return h->values[c];
+}
+
+// bias[n] = (-1<<n) + 1
+static const int stbi__jbias[16] = {0,-1,-3,-7,-15,-31,-63,-127,-255,-511,-1023,-2047,-4095,-8191,-16383,-32767};
+
+// combined JPEG 'receive' and JPEG 'extend', since baseline
+// always extends everything it receives.
+stbi_inline static int stbi__extend_receive(stbi__jpeg *j, int n)
+{
+ unsigned int k;
+ int sgn;
+ if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+
+ sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB
+ k = stbi_lrot(j->code_buffer, n);
+ STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask)));
+ j->code_buffer = k & ~stbi__bmask[n];
+ k &= stbi__bmask[n];
+ j->code_bits -= n;
+ return k + (stbi__jbias[n] & ~sgn);
+}
+
+// get some unsigned bits
+stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n)
+{
+ unsigned int k;
+ if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+ k = stbi_lrot(j->code_buffer, n);
+ j->code_buffer = k & ~stbi__bmask[n];
+ k &= stbi__bmask[n];
+ j->code_bits -= n;
+ return k;
+}
+
+stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j)
+{
+ unsigned int k;
+ if (j->code_bits < 1) stbi__grow_buffer_unsafe(j);
+ k = j->code_buffer;
+ j->code_buffer <<= 1;
+ --j->code_bits;
+ return k & 0x80000000;
+}
+
+// given a value that's at position X in the zigzag stream,
+// where does it appear in the 8x8 matrix coded as row-major?
+static const stbi_uc stbi__jpeg_dezigzag[64+15] =
+{
+ 0, 1, 8, 16, 9, 2, 3, 10,
+ 17, 24, 32, 25, 18, 11, 4, 5,
+ 12, 19, 26, 33, 40, 48, 41, 34,
+ 27, 20, 13, 6, 7, 14, 21, 28,
+ 35, 42, 49, 56, 57, 50, 43, 36,
+ 29, 22, 15, 23, 30, 37, 44, 51,
+ 58, 59, 52, 45, 38, 31, 39, 46,
+ 53, 60, 61, 54, 47, 55, 62, 63,
+ // let corrupt input sample past end
+ 63, 63, 63, 63, 63, 63, 63, 63,
+ 63, 63, 63, 63, 63, 63, 63
+};
+
+// decode one 64-entry block--
+static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant)
+{
+ int diff,dc,k;
+ int t;
+
+ if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+ t = stbi__jpeg_huff_decode(j, hdc);
+ if (t < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+
+ // 0 all the ac values now so we can do it 32-bits at a time
+ memset(data,0,64*sizeof(data[0]));
+
+ diff = t ? stbi__extend_receive(j, t) : 0;
+ dc = j->img_comp[b].dc_pred + diff;
+ j->img_comp[b].dc_pred = dc;
+ data[0] = (short) (dc * dequant[0]);
+
+ // decode AC components, see JPEG spec
+ k = 1;
+ do {
+ unsigned int zig;
+ int c,r,s;
+ if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+ c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+ r = fac[c];
+ if (r) { // fast-AC path
+ k += (r >> 4) & 15; // run
+ s = r & 15; // combined length
+ j->code_buffer <<= s;
+ j->code_bits -= s;
+ // decode into unzigzag'd location
+ zig = stbi__jpeg_dezigzag[k++];
+ data[zig] = (short) ((r >> 8) * dequant[zig]);
+ } else {
+ int rs = stbi__jpeg_huff_decode(j, hac);
+ if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+ s = rs & 15;
+ r = rs >> 4;
+ if (s == 0) {
+ if (rs != 0xf0) break; // end block
+ k += 16;
+ } else {
+ k += r;
+ // decode into unzigzag'd location
+ zig = stbi__jpeg_dezigzag[k++];
+ data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]);
+ }
+ }
+ } while (k < 64);
+ return 1;
+}
+
+static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b)
+{
+ int diff,dc;
+ int t;
+ if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+ if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+ if (j->succ_high == 0) {
+ // first scan for DC coefficient, must be first
+ memset(data,0,64*sizeof(data[0])); // 0 all the ac values now
+ t = stbi__jpeg_huff_decode(j, hdc);
+ diff = t ? stbi__extend_receive(j, t) : 0;
+
+ dc = j->img_comp[b].dc_pred + diff;
+ j->img_comp[b].dc_pred = dc;
+ data[0] = (short) (dc << j->succ_low);
+ } else {
+ // refinement scan for DC coefficient
+ if (stbi__jpeg_get_bit(j))
+ data[0] += (short) (1 << j->succ_low);
+ }
+ return 1;
+}
+
+// @OPTIMIZE: store non-zigzagged during the decode passes,
+// and only de-zigzag when dequantizing
+static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac)
+{
+ int k;
+ if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+ if (j->succ_high == 0) {
+ int shift = j->succ_low;
+
+ if (j->eob_run) {
+ --j->eob_run;
+ return 1;
+ }
+
+ k = j->spec_start;
+ do {
+ unsigned int zig;
+ int c,r,s;
+ if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+ c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+ r = fac[c];
+ if (r) { // fast-AC path
+ k += (r >> 4) & 15; // run
+ s = r & 15; // combined length
+ j->code_buffer <<= s;
+ j->code_bits -= s;
+ zig = stbi__jpeg_dezigzag[k++];
+ data[zig] = (short) ((r >> 8) << shift);
+ } else {
+ int rs = stbi__jpeg_huff_decode(j, hac);
+ if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+ s = rs & 15;
+ r = rs >> 4;
+ if (s == 0) {
+ if (r < 15) {
+ j->eob_run = (1 << r);
+ if (r)
+ j->eob_run += stbi__jpeg_get_bits(j, r);
+ --j->eob_run;
+ break;
+ }
+ k += 16;
+ } else {
+ k += r;
+ zig = stbi__jpeg_dezigzag[k++];
+ data[zig] = (short) (stbi__extend_receive(j,s) << shift);
+ }
+ }
+ } while (k <= j->spec_end);
+ } else {
+ // refinement scan for these AC coefficients
+
+ short bit = (short) (1 << j->succ_low);
+
+ if (j->eob_run) {
+ --j->eob_run;
+ for (k = j->spec_start; k <= j->spec_end; ++k) {
+ short *p = &data[stbi__jpeg_dezigzag[k]];
+ if (*p != 0)
+ if (stbi__jpeg_get_bit(j))
+ if ((*p & bit)==0) {
+ if (*p > 0)
+ *p += bit;
+ else
+ *p -= bit;
+ }
+ }
+ } else {
+ k = j->spec_start;
+ do {
+ int r,s;
+ int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh
+ if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+ s = rs & 15;
+ r = rs >> 4;
+ if (s == 0) {
+ if (r < 15) {
+ j->eob_run = (1 << r) - 1;
+ if (r)
+ j->eob_run += stbi__jpeg_get_bits(j, r);
+ r = 64; // force end of block
+ } else {
+ // r=15 s=0 should write 16 0s, so we just do
+ // a run of 15 0s and then write s (which is 0),
+ // so we don't have to do anything special here
+ }
+ } else {
+ if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG");
+ // sign bit
+ if (stbi__jpeg_get_bit(j))
+ s = bit;
+ else
+ s = -bit;
+ }
+
+ // advance by r
+ while (k <= j->spec_end) {
+ short *p = &data[stbi__jpeg_dezigzag[k++]];
+ if (*p != 0) {
+ if (stbi__jpeg_get_bit(j))
+ if ((*p & bit)==0) {
+ if (*p > 0)
+ *p += bit;
+ else
+ *p -= bit;
+ }
+ } else {
+ if (r == 0) {
+ *p = (short) s;
+ break;
+ }
+ --r;
+ }
+ }
+ } while (k <= j->spec_end);
+ }
+ }
+ return 1;
+}
+
+// take a -128..127 value and stbi__clamp it and convert to 0..255
+stbi_inline static stbi_uc stbi__clamp(int x)
+{
+ // trick to use a single test to catch both cases
+ if ((unsigned int) x > 255) {
+ if (x < 0) return 0;
+ if (x > 255) return 255;
+ }
+ return (stbi_uc) x;
+}
+
+#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5)))
+#define stbi__fsh(x) ((x) * 4096)
+
+// derived from jidctint -- DCT_ISLOW
+#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \
+ int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \
+ p2 = s2; \
+ p3 = s6; \
+ p1 = (p2+p3) * stbi__f2f(0.5411961f); \
+ t2 = p1 + p3*stbi__f2f(-1.847759065f); \
+ t3 = p1 + p2*stbi__f2f( 0.765366865f); \
+ p2 = s0; \
+ p3 = s4; \
+ t0 = stbi__fsh(p2+p3); \
+ t1 = stbi__fsh(p2-p3); \
+ x0 = t0+t3; \
+ x3 = t0-t3; \
+ x1 = t1+t2; \
+ x2 = t1-t2; \
+ t0 = s7; \
+ t1 = s5; \
+ t2 = s3; \
+ t3 = s1; \
+ p3 = t0+t2; \
+ p4 = t1+t3; \
+ p1 = t0+t3; \
+ p2 = t1+t2; \
+ p5 = (p3+p4)*stbi__f2f( 1.175875602f); \
+ t0 = t0*stbi__f2f( 0.298631336f); \
+ t1 = t1*stbi__f2f( 2.053119869f); \
+ t2 = t2*stbi__f2f( 3.072711026f); \
+ t3 = t3*stbi__f2f( 1.501321110f); \
+ p1 = p5 + p1*stbi__f2f(-0.899976223f); \
+ p2 = p5 + p2*stbi__f2f(-2.562915447f); \
+ p3 = p3*stbi__f2f(-1.961570560f); \
+ p4 = p4*stbi__f2f(-0.390180644f); \
+ t3 += p1+p4; \
+ t2 += p2+p3; \
+ t1 += p2+p4; \
+ t0 += p1+p3;
+
+static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64])
+{
+ int i,val[64],*v=val;
+ stbi_uc *o;
+ short *d = data;
+
+ // columns
+ for (i=0; i < 8; ++i,++d, ++v) {
+ // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing
+ if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0
+ && d[40]==0 && d[48]==0 && d[56]==0) {
+ // no shortcut 0 seconds
+ // (1|2|3|4|5|6|7)==0 0 seconds
+ // all separate -0.047 seconds
+ // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds
+ int dcterm = d[0]*4;
+ v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm;
+ } else {
+ STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56])
+ // constants scaled things up by 1<<12; let's bring them back
+ // down, but keep 2 extra bits of precision
+ x0 += 512; x1 += 512; x2 += 512; x3 += 512;
+ v[ 0] = (x0+t3) >> 10;
+ v[56] = (x0-t3) >> 10;
+ v[ 8] = (x1+t2) >> 10;
+ v[48] = (x1-t2) >> 10;
+ v[16] = (x2+t1) >> 10;
+ v[40] = (x2-t1) >> 10;
+ v[24] = (x3+t0) >> 10;
+ v[32] = (x3-t0) >> 10;
+ }
+ }
+
+ for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) {
+ // no fast case since the first 1D IDCT spread components out
+ STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7])
+ // constants scaled things up by 1<<12, plus we had 1<<2 from first
+ // loop, plus horizontal and vertical each scale by sqrt(8) so together
+ // we've got an extra 1<<3, so 1<<17 total we need to remove.
+ // so we want to round that, which means adding 0.5 * 1<<17,
+ // aka 65536. Also, we'll end up with -128 to 127 that we want
+ // to encode as 0..255 by adding 128, so we'll add that before the shift
+ x0 += 65536 + (128<<17);
+ x1 += 65536 + (128<<17);
+ x2 += 65536 + (128<<17);
+ x3 += 65536 + (128<<17);
+ // tried computing the shifts into temps, or'ing the temps to see
+ // if any were out of range, but that was slower
+ o[0] = stbi__clamp((x0+t3) >> 17);
+ o[7] = stbi__clamp((x0-t3) >> 17);
+ o[1] = stbi__clamp((x1+t2) >> 17);
+ o[6] = stbi__clamp((x1-t2) >> 17);
+ o[2] = stbi__clamp((x2+t1) >> 17);
+ o[5] = stbi__clamp((x2-t1) >> 17);
+ o[3] = stbi__clamp((x3+t0) >> 17);
+ o[4] = stbi__clamp((x3-t0) >> 17);
+ }
+}
+
+#ifdef STBI_SSE2
+// sse2 integer IDCT. not the fastest possible implementation but it
+// produces bit-identical results to the generic C version so it's
+// fully "transparent".
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+ // This is constructed to match our regular (generic) integer IDCT exactly.
+ __m128i row0, row1, row2, row3, row4, row5, row6, row7;
+ __m128i tmp;
+
+ // dot product constant: even elems=x, odd elems=y
+ #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y))
+
+ // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit)
+ // out(1) = c1[even]*x + c1[odd]*y
+ #define dct_rot(out0,out1, x,y,c0,c1) \
+ __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \
+ __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \
+ __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \
+ __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \
+ __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \
+ __m128i out1##_h = _mm_madd_epi16(c0##hi, c1)
+
+ // out = in << 12 (in 16-bit, out 32-bit)
+ #define dct_widen(out, in) \
+ __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \
+ __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4)
+
+ // wide add
+ #define dct_wadd(out, a, b) \
+ __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \
+ __m128i out##_h = _mm_add_epi32(a##_h, b##_h)
+
+ // wide sub
+ #define dct_wsub(out, a, b) \
+ __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \
+ __m128i out##_h = _mm_sub_epi32(a##_h, b##_h)
+
+ // butterfly a/b, add bias, then shift by "s" and pack
+ #define dct_bfly32o(out0, out1, a,b,bias,s) \
+ { \
+ __m128i abiased_l = _mm_add_epi32(a##_l, bias); \
+ __m128i abiased_h = _mm_add_epi32(a##_h, bias); \
+ dct_wadd(sum, abiased, b); \
+ dct_wsub(dif, abiased, b); \
+ out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \
+ out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \
+ }
+
+ // 8-bit interleave step (for transposes)
+ #define dct_interleave8(a, b) \
+ tmp = a; \
+ a = _mm_unpacklo_epi8(a, b); \
+ b = _mm_unpackhi_epi8(tmp, b)
+
+ // 16-bit interleave step (for transposes)
+ #define dct_interleave16(a, b) \
+ tmp = a; \
+ a = _mm_unpacklo_epi16(a, b); \
+ b = _mm_unpackhi_epi16(tmp, b)
+
+ #define dct_pass(bias,shift) \
+ { \
+ /* even part */ \
+ dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \
+ __m128i sum04 = _mm_add_epi16(row0, row4); \
+ __m128i dif04 = _mm_sub_epi16(row0, row4); \
+ dct_widen(t0e, sum04); \
+ dct_widen(t1e, dif04); \
+ dct_wadd(x0, t0e, t3e); \
+ dct_wsub(x3, t0e, t3e); \
+ dct_wadd(x1, t1e, t2e); \
+ dct_wsub(x2, t1e, t2e); \
+ /* odd part */ \
+ dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \
+ dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \
+ __m128i sum17 = _mm_add_epi16(row1, row7); \
+ __m128i sum35 = _mm_add_epi16(row3, row5); \
+ dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \
+ dct_wadd(x4, y0o, y4o); \
+ dct_wadd(x5, y1o, y5o); \
+ dct_wadd(x6, y2o, y5o); \
+ dct_wadd(x7, y3o, y4o); \
+ dct_bfly32o(row0,row7, x0,x7,bias,shift); \
+ dct_bfly32o(row1,row6, x1,x6,bias,shift); \
+ dct_bfly32o(row2,row5, x2,x5,bias,shift); \
+ dct_bfly32o(row3,row4, x3,x4,bias,shift); \
+ }
+
+ __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f));
+ __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f));
+ __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f));
+ __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f));
+ __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f));
+ __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f));
+ __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f));
+ __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f));
+
+ // rounding biases in column/row passes, see stbi__idct_block for explanation.
+ __m128i bias_0 = _mm_set1_epi32(512);
+ __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17));
+
+ // load
+ row0 = _mm_load_si128((const __m128i *) (data + 0*8));
+ row1 = _mm_load_si128((const __m128i *) (data + 1*8));
+ row2 = _mm_load_si128((const __m128i *) (data + 2*8));
+ row3 = _mm_load_si128((const __m128i *) (data + 3*8));
+ row4 = _mm_load_si128((const __m128i *) (data + 4*8));
+ row5 = _mm_load_si128((const __m128i *) (data + 5*8));
+ row6 = _mm_load_si128((const __m128i *) (data + 6*8));
+ row7 = _mm_load_si128((const __m128i *) (data + 7*8));
+
+ // column pass
+ dct_pass(bias_0, 10);
+
+ {
+ // 16bit 8x8 transpose pass 1
+ dct_interleave16(row0, row4);
+ dct_interleave16(row1, row5);
+ dct_interleave16(row2, row6);
+ dct_interleave16(row3, row7);
+
+ // transpose pass 2
+ dct_interleave16(row0, row2);
+ dct_interleave16(row1, row3);
+ dct_interleave16(row4, row6);
+ dct_interleave16(row5, row7);
+
+ // transpose pass 3
+ dct_interleave16(row0, row1);
+ dct_interleave16(row2, row3);
+ dct_interleave16(row4, row5);
+ dct_interleave16(row6, row7);
+ }
+
+ // row pass
+ dct_pass(bias_1, 17);
+
+ {
+ // pack
+ __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7
+ __m128i p1 = _mm_packus_epi16(row2, row3);
+ __m128i p2 = _mm_packus_epi16(row4, row5);
+ __m128i p3 = _mm_packus_epi16(row6, row7);
+
+ // 8bit 8x8 transpose pass 1
+ dct_interleave8(p0, p2); // a0e0a1e1...
+ dct_interleave8(p1, p3); // c0g0c1g1...
+
+ // transpose pass 2
+ dct_interleave8(p0, p1); // a0c0e0g0...
+ dct_interleave8(p2, p3); // b0d0f0h0...
+
+ // transpose pass 3
+ dct_interleave8(p0, p2); // a0b0c0d0...
+ dct_interleave8(p1, p3); // a4b4c4d4...
+
+ // store
+ _mm_storel_epi64((__m128i *) out, p0); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, p2); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, p1); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, p3); out += out_stride;
+ _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e));
+ }
+
+#undef dct_const
+#undef dct_rot
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_interleave8
+#undef dct_interleave16
+#undef dct_pass
+}
+
+#endif // STBI_SSE2
+
+#ifdef STBI_NEON
+
+// NEON integer IDCT. should produce bit-identical
+// results to the generic C version.
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+ int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;
+
+ int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f));
+ int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f));
+ int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f));
+ int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f));
+ int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f));
+ int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f));
+ int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f));
+ int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f));
+ int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f));
+ int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f));
+ int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f));
+ int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f));
+
+#define dct_long_mul(out, inq, coeff) \
+ int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \
+ int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff)
+
+#define dct_long_mac(out, acc, inq, coeff) \
+ int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \
+ int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff)
+
+#define dct_widen(out, inq) \
+ int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \
+ int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12)
+
+// wide add
+#define dct_wadd(out, a, b) \
+ int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \
+ int32x4_t out##_h = vaddq_s32(a##_h, b##_h)
+
+// wide sub
+#define dct_wsub(out, a, b) \
+ int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \
+ int32x4_t out##_h = vsubq_s32(a##_h, b##_h)
+
+// butterfly a/b, then shift using "shiftop" by "s" and pack
+#define dct_bfly32o(out0,out1, a,b,shiftop,s) \
+ { \
+ dct_wadd(sum, a, b); \
+ dct_wsub(dif, a, b); \
+ out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \
+ out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \
+ }
+
+#define dct_pass(shiftop, shift) \
+ { \
+ /* even part */ \
+ int16x8_t sum26 = vaddq_s16(row2, row6); \
+ dct_long_mul(p1e, sum26, rot0_0); \
+ dct_long_mac(t2e, p1e, row6, rot0_1); \
+ dct_long_mac(t3e, p1e, row2, rot0_2); \
+ int16x8_t sum04 = vaddq_s16(row0, row4); \
+ int16x8_t dif04 = vsubq_s16(row0, row4); \
+ dct_widen(t0e, sum04); \
+ dct_widen(t1e, dif04); \
+ dct_wadd(x0, t0e, t3e); \
+ dct_wsub(x3, t0e, t3e); \
+ dct_wadd(x1, t1e, t2e); \
+ dct_wsub(x2, t1e, t2e); \
+ /* odd part */ \
+ int16x8_t sum15 = vaddq_s16(row1, row5); \
+ int16x8_t sum17 = vaddq_s16(row1, row7); \
+ int16x8_t sum35 = vaddq_s16(row3, row5); \
+ int16x8_t sum37 = vaddq_s16(row3, row7); \
+ int16x8_t sumodd = vaddq_s16(sum17, sum35); \
+ dct_long_mul(p5o, sumodd, rot1_0); \
+ dct_long_mac(p1o, p5o, sum17, rot1_1); \
+ dct_long_mac(p2o, p5o, sum35, rot1_2); \
+ dct_long_mul(p3o, sum37, rot2_0); \
+ dct_long_mul(p4o, sum15, rot2_1); \
+ dct_wadd(sump13o, p1o, p3o); \
+ dct_wadd(sump24o, p2o, p4o); \
+ dct_wadd(sump23o, p2o, p3o); \
+ dct_wadd(sump14o, p1o, p4o); \
+ dct_long_mac(x4, sump13o, row7, rot3_0); \
+ dct_long_mac(x5, sump24o, row5, rot3_1); \
+ dct_long_mac(x6, sump23o, row3, rot3_2); \
+ dct_long_mac(x7, sump14o, row1, rot3_3); \
+ dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \
+ dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \
+ dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \
+ dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \
+ }
+
+ // load
+ row0 = vld1q_s16(data + 0*8);
+ row1 = vld1q_s16(data + 1*8);
+ row2 = vld1q_s16(data + 2*8);
+ row3 = vld1q_s16(data + 3*8);
+ row4 = vld1q_s16(data + 4*8);
+ row5 = vld1q_s16(data + 5*8);
+ row6 = vld1q_s16(data + 6*8);
+ row7 = vld1q_s16(data + 7*8);
+
+ // add DC bias
+ row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0));
+
+ // column pass
+ dct_pass(vrshrn_n_s32, 10);
+
+ // 16bit 8x8 transpose
+ {
+// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively.
+// whether compilers actually get this is another story, sadly.
+#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); }
+#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); }
+
+ // pass 1
+ dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6
+ dct_trn16(row2, row3);
+ dct_trn16(row4, row5);
+ dct_trn16(row6, row7);
+
+ // pass 2
+ dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4
+ dct_trn32(row1, row3);
+ dct_trn32(row4, row6);
+ dct_trn32(row5, row7);
+
+ // pass 3
+ dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0
+ dct_trn64(row1, row5);
+ dct_trn64(row2, row6);
+ dct_trn64(row3, row7);
+
+#undef dct_trn16
+#undef dct_trn32
+#undef dct_trn64
+ }
+
+ // row pass
+ // vrshrn_n_s32 only supports shifts up to 16, we need
+ // 17. so do a non-rounding shift of 16 first then follow
+ // up with a rounding shift by 1.
+ dct_pass(vshrn_n_s32, 16);
+
+ {
+ // pack and round
+ uint8x8_t p0 = vqrshrun_n_s16(row0, 1);
+ uint8x8_t p1 = vqrshrun_n_s16(row1, 1);
+ uint8x8_t p2 = vqrshrun_n_s16(row2, 1);
+ uint8x8_t p3 = vqrshrun_n_s16(row3, 1);
+ uint8x8_t p4 = vqrshrun_n_s16(row4, 1);
+ uint8x8_t p5 = vqrshrun_n_s16(row5, 1);
+ uint8x8_t p6 = vqrshrun_n_s16(row6, 1);
+ uint8x8_t p7 = vqrshrun_n_s16(row7, 1);
+
+ // again, these can translate into one instruction, but often don't.
+#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); }
+#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); }
+
+ // sadly can't use interleaved stores here since we only write
+ // 8 bytes to each scan line!
+
+ // 8x8 8-bit transpose pass 1
+ dct_trn8_8(p0, p1);
+ dct_trn8_8(p2, p3);
+ dct_trn8_8(p4, p5);
+ dct_trn8_8(p6, p7);
+
+ // pass 2
+ dct_trn8_16(p0, p2);
+ dct_trn8_16(p1, p3);
+ dct_trn8_16(p4, p6);
+ dct_trn8_16(p5, p7);
+
+ // pass 3
+ dct_trn8_32(p0, p4);
+ dct_trn8_32(p1, p5);
+ dct_trn8_32(p2, p6);
+ dct_trn8_32(p3, p7);
+
+ // store
+ vst1_u8(out, p0); out += out_stride;
+ vst1_u8(out, p1); out += out_stride;
+ vst1_u8(out, p2); out += out_stride;
+ vst1_u8(out, p3); out += out_stride;
+ vst1_u8(out, p4); out += out_stride;
+ vst1_u8(out, p5); out += out_stride;
+ vst1_u8(out, p6); out += out_stride;
+ vst1_u8(out, p7);
+
+#undef dct_trn8_8
+#undef dct_trn8_16
+#undef dct_trn8_32
+ }
+
+#undef dct_long_mul
+#undef dct_long_mac
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_pass
+}
+
+#endif // STBI_NEON
+
+#define STBI__MARKER_none 0xff
+// if there's a pending marker from the entropy stream, return that
+// otherwise, fetch from the stream and get a marker. if there's no
+// marker, return 0xff, which is never a valid marker value
+static stbi_uc stbi__get_marker(stbi__jpeg *j)
+{
+ stbi_uc x;
+ if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; }
+ x = stbi__get8(j->s);
+ if (x != 0xff) return STBI__MARKER_none;
+ while (x == 0xff)
+ x = stbi__get8(j->s); // consume repeated 0xff fill bytes
+ return x;
+}
+
+// in each scan, we'll have scan_n components, and the order
+// of the components is specified by order[]
+#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7)
+
+// after a restart interval, stbi__jpeg_reset the entropy decoder and
+// the dc prediction
+static void stbi__jpeg_reset(stbi__jpeg *j)
+{
+ j->code_bits = 0;
+ j->code_buffer = 0;
+ j->nomore = 0;
+ j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0;
+ j->marker = STBI__MARKER_none;
+ j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff;
+ j->eob_run = 0;
+ // no more than 1<<31 MCUs if no restart_interal? that's plenty safe,
+ // since we don't even allow 1<<30 pixels
+}
+
+static int stbi__parse_entropy_coded_data(stbi__jpeg *z)
+{
+ stbi__jpeg_reset(z);
+ if (!z->progressive) {
+ if (z->scan_n == 1) {
+ int i,j;
+ STBI_SIMD_ALIGN(short, data[64]);
+ int n = z->order[0];
+ // non-interleaved data, we just need to process one block at a time,
+ // in trivial scanline order
+ // number of blocks to do just depends on how many actual "pixels" this
+ // component has, independent of interleaved MCU blocking and such
+ int w = (z->img_comp[n].x+7) >> 3;
+ int h = (z->img_comp[n].y+7) >> 3;
+ for (j=0; j < h; ++j) {
+ for (i=0; i < w; ++i) {
+ int ha = z->img_comp[n].ha;
+ if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+ z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+ // every data block is an MCU, so countdown the restart interval
+ if (--z->todo <= 0) {
+ if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+ // if it's NOT a restart, then just bail, so we get corrupt data
+ // rather than no data
+ if (!STBI__RESTART(z->marker)) return 1;
+ stbi__jpeg_reset(z);
+ }
+ }
+ }
+ return 1;
+ } else { // interleaved
+ int i,j,k,x,y;
+ STBI_SIMD_ALIGN(short, data[64]);
+ for (j=0; j < z->img_mcu_y; ++j) {
+ for (i=0; i < z->img_mcu_x; ++i) {
+ // scan an interleaved mcu... process scan_n components in order
+ for (k=0; k < z->scan_n; ++k) {
+ int n = z->order[k];
+ // scan out an mcu's worth of this component; that's just determined
+ // by the basic H and V specified for the component
+ for (y=0; y < z->img_comp[n].v; ++y) {
+ for (x=0; x < z->img_comp[n].h; ++x) {
+ int x2 = (i*z->img_comp[n].h + x)*8;
+ int y2 = (j*z->img_comp[n].v + y)*8;
+ int ha = z->img_comp[n].ha;
+ if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+ z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data);
+ }
+ }
+ }
+ // after all interleaved components, that's an interleaved MCU,
+ // so now count down the restart interval
+ if (--z->todo <= 0) {
+ if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+ if (!STBI__RESTART(z->marker)) return 1;
+ stbi__jpeg_reset(z);
+ }
+ }
+ }
+ return 1;
+ }
+ } else {
+ if (z->scan_n == 1) {
+ int i,j;
+ int n = z->order[0];
+ // non-interleaved data, we just need to process one block at a time,
+ // in trivial scanline order
+ // number of blocks to do just depends on how many actual "pixels" this
+ // component has, independent of interleaved MCU blocking and such
+ int w = (z->img_comp[n].x+7) >> 3;
+ int h = (z->img_comp[n].y+7) >> 3;
+ for (j=0; j < h; ++j) {
+ for (i=0; i < w; ++i) {
+ short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+ if (z->spec_start == 0) {
+ if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+ return 0;
+ } else {
+ int ha = z->img_comp[n].ha;
+ if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha]))
+ return 0;
+ }
+ // every data block is an MCU, so countdown the restart interval
+ if (--z->todo <= 0) {
+ if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+ if (!STBI__RESTART(z->marker)) return 1;
+ stbi__jpeg_reset(z);
+ }
+ }
+ }
+ return 1;
+ } else { // interleaved
+ int i,j,k,x,y;
+ for (j=0; j < z->img_mcu_y; ++j) {
+ for (i=0; i < z->img_mcu_x; ++i) {
+ // scan an interleaved mcu... process scan_n components in order
+ for (k=0; k < z->scan_n; ++k) {
+ int n = z->order[k];
+ // scan out an mcu's worth of this component; that's just determined
+ // by the basic H and V specified for the component
+ for (y=0; y < z->img_comp[n].v; ++y) {
+ for (x=0; x < z->img_comp[n].h; ++x) {
+ int x2 = (i*z->img_comp[n].h + x);
+ int y2 = (j*z->img_comp[n].v + y);
+ short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w);
+ if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+ return 0;
+ }
+ }
+ }
+ // after all interleaved components, that's an interleaved MCU,
+ // so now count down the restart interval
+ if (--z->todo <= 0) {
+ if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+ if (!STBI__RESTART(z->marker)) return 1;
+ stbi__jpeg_reset(z);
+ }
+ }
+ }
+ return 1;
+ }
+ }
+}
+
+static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant)
+{
+ int i;
+ for (i=0; i < 64; ++i)
+ data[i] *= dequant[i];
+}
+
+static void stbi__jpeg_finish(stbi__jpeg *z)
+{
+ if (z->progressive) {
+ // dequantize and idct the data
+ int i,j,n;
+ for (n=0; n < z->s->img_n; ++n) {
+ int w = (z->img_comp[n].x+7) >> 3;
+ int h = (z->img_comp[n].y+7) >> 3;
+ for (j=0; j < h; ++j) {
+ for (i=0; i < w; ++i) {
+ short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+ stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]);
+ z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+ }
+ }
+ }
+ }
+}
+
+static int stbi__process_marker(stbi__jpeg *z, int m)
+{
+ int L;
+ switch (m) {
+ case STBI__MARKER_none: // no marker found
+ return stbi__err("expected marker","Corrupt JPEG");
+
+ case 0xDD: // DRI - specify restart interval
+ if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG");
+ z->restart_interval = stbi__get16be(z->s);
+ return 1;
+
+ case 0xDB: // DQT - define quantization table
+ L = stbi__get16be(z->s)-2;
+ while (L > 0) {
+ int q = stbi__get8(z->s);
+ int p = q >> 4, sixteen = (p != 0);
+ int t = q & 15,i;
+ if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG");
+ if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG");
+
+ for (i=0; i < 64; ++i)
+ z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s));
+ L -= (sixteen ? 129 : 65);
+ }
+ return L==0;
+
+ case 0xC4: // DHT - define huffman table
+ L = stbi__get16be(z->s)-2;
+ while (L > 0) {
+ stbi_uc *v;
+ int sizes[16],i,n=0;
+ int q = stbi__get8(z->s);
+ int tc = q >> 4;
+ int th = q & 15;
+ if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG");
+ for (i=0; i < 16; ++i) {
+ sizes[i] = stbi__get8(z->s);
+ n += sizes[i];
+ }
+ L -= 17;
+ if (tc == 0) {
+ if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0;
+ v = z->huff_dc[th].values;
+ } else {
+ if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0;
+ v = z->huff_ac[th].values;
+ }
+ for (i=0; i < n; ++i)
+ v[i] = stbi__get8(z->s);
+ if (tc != 0)
+ stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th);
+ L -= n;
+ }
+ return L==0;
+ }
+
+ // check for comment block or APP blocks
+ if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) {
+ L = stbi__get16be(z->s);
+ if (L < 2) {
+ if (m == 0xFE)
+ return stbi__err("bad COM len","Corrupt JPEG");
+ else
+ return stbi__err("bad APP len","Corrupt JPEG");
+ }
+ L -= 2;
+
+ if (m == 0xE0 && L >= 5) { // JFIF APP0 segment
+ static const unsigned char tag[5] = {'J','F','I','F','\0'};
+ int ok = 1;
+ int i;
+ for (i=0; i < 5; ++i)
+ if (stbi__get8(z->s) != tag[i])
+ ok = 0;
+ L -= 5;
+ if (ok)
+ z->jfif = 1;
+ } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment
+ static const unsigned char tag[6] = {'A','d','o','b','e','\0'};
+ int ok = 1;
+ int i;
+ for (i=0; i < 6; ++i)
+ if (stbi__get8(z->s) != tag[i])
+ ok = 0;
+ L -= 6;
+ if (ok) {
+ stbi__get8(z->s); // version
+ stbi__get16be(z->s); // flags0
+ stbi__get16be(z->s); // flags1
+ z->app14_color_transform = stbi__get8(z->s); // color transform
+ L -= 6;
+ }
+ }
+
+ stbi__skip(z->s, L);
+ return 1;
+ }
+
+ return stbi__err("unknown marker","Corrupt JPEG");
+}
+
+// after we see SOS
+static int stbi__process_scan_header(stbi__jpeg *z)
+{
+ int i;
+ int Ls = stbi__get16be(z->s);
+ z->scan_n = stbi__get8(z->s);
+ if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG");
+ if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG");
+ for (i=0; i < z->scan_n; ++i) {
+ int id = stbi__get8(z->s), which;
+ int q = stbi__get8(z->s);
+ for (which = 0; which < z->s->img_n; ++which)
+ if (z->img_comp[which].id == id)
+ break;
+ if (which == z->s->img_n) return 0; // no match
+ z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG");
+ z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG");
+ z->order[i] = which;
+ }
+
+ {
+ int aa;
+ z->spec_start = stbi__get8(z->s);
+ z->spec_end = stbi__get8(z->s); // should be 63, but might be 0
+ aa = stbi__get8(z->s);
+ z->succ_high = (aa >> 4);
+ z->succ_low = (aa & 15);
+ if (z->progressive) {
+ if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13)
+ return stbi__err("bad SOS", "Corrupt JPEG");
+ } else {
+ if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG");
+ if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG");
+ z->spec_end = 63;
+ }
+ }
+
+ return 1;
+}
+
+static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why)
+{
+ int i;
+ for (i=0; i < ncomp; ++i) {
+ if (z->img_comp[i].raw_data) {
+ STBI_FREE(z->img_comp[i].raw_data);
+ z->img_comp[i].raw_data = NULL;
+ z->img_comp[i].data = NULL;
+ }
+ if (z->img_comp[i].raw_coeff) {
+ STBI_FREE(z->img_comp[i].raw_coeff);
+ z->img_comp[i].raw_coeff = 0;
+ z->img_comp[i].coeff = 0;
+ }
+ if (z->img_comp[i].linebuf) {
+ STBI_FREE(z->img_comp[i].linebuf);
+ z->img_comp[i].linebuf = NULL;
+ }
+ }
+ return why;
+}
+
+static int stbi__process_frame_header(stbi__jpeg *z, int scan)
+{
+ stbi__context *s = z->s;
+ int Lf,p,i,q, h_max=1,v_max=1,c;
+ Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG
+ p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline
+ s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG
+ s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires
+ c = stbi__get8(s);
+ if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG");
+ s->img_n = c;
+ for (i=0; i < c; ++i) {
+ z->img_comp[i].data = NULL;
+ z->img_comp[i].linebuf = NULL;
+ }
+
+ if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG");
+
+ z->rgb = 0;
+ for (i=0; i < s->img_n; ++i) {
+ static const unsigned char rgb[3] = { 'R', 'G', 'B' };
+ z->img_comp[i].id = stbi__get8(s);
+ if (s->img_n == 3 && z->img_comp[i].id == rgb[i])
+ ++z->rgb;
+ q = stbi__get8(s);
+ z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG");
+ z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG");
+ z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG");
+ }
+
+ if (scan != STBI__SCAN_load) return 1;
+
+ if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode");
+
+ for (i=0; i < s->img_n; ++i) {
+ if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h;
+ if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v;
+ }
+
+ // compute interleaved mcu info
+ z->img_h_max = h_max;
+ z->img_v_max = v_max;
+ z->img_mcu_w = h_max * 8;
+ z->img_mcu_h = v_max * 8;
+ // these sizes can't be more than 17 bits
+ z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w;
+ z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h;
+
+ for (i=0; i < s->img_n; ++i) {
+ // number of effective pixels (e.g. for non-interleaved MCU)
+ z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max;
+ z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max;
+ // to simplify generation, we'll allocate enough memory to decode
+ // the bogus oversized data from using interleaved MCUs and their
+ // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't
+ // discard the extra data until colorspace conversion
+ //
+ // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier)
+ // so these muls can't overflow with 32-bit ints (which we require)
+ z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8;
+ z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8;
+ z->img_comp[i].coeff = 0;
+ z->img_comp[i].raw_coeff = 0;
+ z->img_comp[i].linebuf = NULL;
+ z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15);
+ if (z->img_comp[i].raw_data == NULL)
+ return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+ // align blocks for idct using mmx/sse
+ z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15);
+ if (z->progressive) {
+ // w2, h2 are multiples of 8 (see above)
+ z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8;
+ z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8;
+ z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15);
+ if (z->img_comp[i].raw_coeff == NULL)
+ return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+ z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15);
+ }
+ }
+
+ return 1;
+}
+
+// use comparisons since in some cases we handle more than one case (e.g. SOF)
+#define stbi__DNL(x) ((x) == 0xdc)
+#define stbi__SOI(x) ((x) == 0xd8)
+#define stbi__EOI(x) ((x) == 0xd9)
+#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2)
+#define stbi__SOS(x) ((x) == 0xda)
+
+#define stbi__SOF_progressive(x) ((x) == 0xc2)
+
+static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan)
+{
+ int m;
+ z->jfif = 0;
+ z->app14_color_transform = -1; // valid values are 0,1,2
+ z->marker = STBI__MARKER_none; // initialize cached marker to empty
+ m = stbi__get_marker(z);
+ if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG");
+ if (scan == STBI__SCAN_type) return 1;
+ m = stbi__get_marker(z);
+ while (!stbi__SOF(m)) {
+ if (!stbi__process_marker(z,m)) return 0;
+ m = stbi__get_marker(z);
+ while (m == STBI__MARKER_none) {
+ // some files have extra padding after their blocks, so ok, we'll scan
+ if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG");
+ m = stbi__get_marker(z);
+ }
+ }
+ z->progressive = stbi__SOF_progressive(m);
+ if (!stbi__process_frame_header(z, scan)) return 0;
+ return 1;
+}
+
+// decode image to YCbCr format
+static int stbi__decode_jpeg_image(stbi__jpeg *j)
+{
+ int m;
+ for (m = 0; m < 4; m++) {
+ j->img_comp[m].raw_data = NULL;
+ j->img_comp[m].raw_coeff = NULL;
+ }
+ j->restart_interval = 0;
+ if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0;
+ m = stbi__get_marker(j);
+ while (!stbi__EOI(m)) {
+ if (stbi__SOS(m)) {
+ if (!stbi__process_scan_header(j)) return 0;
+ if (!stbi__parse_entropy_coded_data(j)) return 0;
+ if (j->marker == STBI__MARKER_none ) {
+ // handle 0s at the end of image data from IP Kamera 9060
+ while (!stbi__at_eof(j->s)) {
+ int x = stbi__get8(j->s);
+ if (x == 255) {
+ j->marker = stbi__get8(j->s);
+ break;
+ }
+ }
+ // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0
+ }
+ } else if (stbi__DNL(m)) {
+ int Ld = stbi__get16be(j->s);
+ stbi__uint32 NL = stbi__get16be(j->s);
+ if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG");
+ if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG");
+ } else {
+ if (!stbi__process_marker(j, m)) return 0;
+ }
+ m = stbi__get_marker(j);
+ }
+ if (j->progressive)
+ stbi__jpeg_finish(j);
+ return 1;
+}
+
+// static jfif-centered resampling (across block boundaries)
+
+typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1,
+ int w, int hs);
+
+#define stbi__div4(x) ((stbi_uc) ((x) >> 2))
+
+static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ STBI_NOTUSED(out);
+ STBI_NOTUSED(in_far);
+ STBI_NOTUSED(w);
+ STBI_NOTUSED(hs);
+ return in_near;
+}
+
+static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ // need to generate two samples vertically for every one in input
+ int i;
+ STBI_NOTUSED(hs);
+ for (i=0; i < w; ++i)
+ out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2);
+ return out;
+}
+
+static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ // need to generate two samples horizontally for every one in input
+ int i;
+ stbi_uc *input = in_near;
+
+ if (w == 1) {
+ // if only one sample, can't do any interpolation
+ out[0] = out[1] = input[0];
+ return out;
+ }
+
+ out[0] = input[0];
+ out[1] = stbi__div4(input[0]*3 + input[1] + 2);
+ for (i=1; i < w-1; ++i) {
+ int n = 3*input[i]+2;
+ out[i*2+0] = stbi__div4(n+input[i-1]);
+ out[i*2+1] = stbi__div4(n+input[i+1]);
+ }
+ out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2);
+ out[i*2+1] = input[w-1];
+
+ STBI_NOTUSED(in_far);
+ STBI_NOTUSED(hs);
+
+ return out;
+}
+
+#define stbi__div16(x) ((stbi_uc) ((x) >> 4))
+
+static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ // need to generate 2x2 samples for every one in input
+ int i,t0,t1;
+ if (w == 1) {
+ out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+ return out;
+ }
+
+ t1 = 3*in_near[0] + in_far[0];
+ out[0] = stbi__div4(t1+2);
+ for (i=1; i < w; ++i) {
+ t0 = t1;
+ t1 = 3*in_near[i]+in_far[i];
+ out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+ out[i*2 ] = stbi__div16(3*t1 + t0 + 8);
+ }
+ out[w*2-1] = stbi__div4(t1+2);
+
+ STBI_NOTUSED(hs);
+
+ return out;
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ // need to generate 2x2 samples for every one in input
+ int i=0,t0,t1;
+
+ if (w == 1) {
+ out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+ return out;
+ }
+
+ t1 = 3*in_near[0] + in_far[0];
+ // process groups of 8 pixels for as long as we can.
+ // note we can't handle the last pixel in a row in this loop
+ // because we need to handle the filter boundary conditions.
+ for (; i < ((w-1) & ~7); i += 8) {
+#if defined(STBI_SSE2)
+ // load and perform the vertical filtering pass
+ // this uses 3*x + y = 4*x + (y - x)
+ __m128i zero = _mm_setzero_si128();
+ __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i));
+ __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i));
+ __m128i farw = _mm_unpacklo_epi8(farb, zero);
+ __m128i nearw = _mm_unpacklo_epi8(nearb, zero);
+ __m128i diff = _mm_sub_epi16(farw, nearw);
+ __m128i nears = _mm_slli_epi16(nearw, 2);
+ __m128i curr = _mm_add_epi16(nears, diff); // current row
+
+ // horizontal filter works the same based on shifted vers of current
+ // row. "prev" is current row shifted right by 1 pixel; we need to
+ // insert the previous pixel value (from t1).
+ // "next" is current row shifted left by 1 pixel, with first pixel
+ // of next block of 8 pixels added in.
+ __m128i prv0 = _mm_slli_si128(curr, 2);
+ __m128i nxt0 = _mm_srli_si128(curr, 2);
+ __m128i prev = _mm_insert_epi16(prv0, t1, 0);
+ __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7);
+
+ // horizontal filter, polyphase implementation since it's convenient:
+ // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+ // odd pixels = 3*cur + next = cur*4 + (next - cur)
+ // note the shared term.
+ __m128i bias = _mm_set1_epi16(8);
+ __m128i curs = _mm_slli_epi16(curr, 2);
+ __m128i prvd = _mm_sub_epi16(prev, curr);
+ __m128i nxtd = _mm_sub_epi16(next, curr);
+ __m128i curb = _mm_add_epi16(curs, bias);
+ __m128i even = _mm_add_epi16(prvd, curb);
+ __m128i odd = _mm_add_epi16(nxtd, curb);
+
+ // interleave even and odd pixels, then undo scaling.
+ __m128i int0 = _mm_unpacklo_epi16(even, odd);
+ __m128i int1 = _mm_unpackhi_epi16(even, odd);
+ __m128i de0 = _mm_srli_epi16(int0, 4);
+ __m128i de1 = _mm_srli_epi16(int1, 4);
+
+ // pack and write output
+ __m128i outv = _mm_packus_epi16(de0, de1);
+ _mm_storeu_si128((__m128i *) (out + i*2), outv);
+#elif defined(STBI_NEON)
+ // load and perform the vertical filtering pass
+ // this uses 3*x + y = 4*x + (y - x)
+ uint8x8_t farb = vld1_u8(in_far + i);
+ uint8x8_t nearb = vld1_u8(in_near + i);
+ int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb));
+ int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2));
+ int16x8_t curr = vaddq_s16(nears, diff); // current row
+
+ // horizontal filter works the same based on shifted vers of current
+ // row. "prev" is current row shifted right by 1 pixel; we need to
+ // insert the previous pixel value (from t1).
+ // "next" is current row shifted left by 1 pixel, with first pixel
+ // of next block of 8 pixels added in.
+ int16x8_t prv0 = vextq_s16(curr, curr, 7);
+ int16x8_t nxt0 = vextq_s16(curr, curr, 1);
+ int16x8_t prev = vsetq_lane_s16(t1, prv0, 0);
+ int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7);
+
+ // horizontal filter, polyphase implementation since it's convenient:
+ // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+ // odd pixels = 3*cur + next = cur*4 + (next - cur)
+ // note the shared term.
+ int16x8_t curs = vshlq_n_s16(curr, 2);
+ int16x8_t prvd = vsubq_s16(prev, curr);
+ int16x8_t nxtd = vsubq_s16(next, curr);
+ int16x8_t even = vaddq_s16(curs, prvd);
+ int16x8_t odd = vaddq_s16(curs, nxtd);
+
+ // undo scaling and round, then store with even/odd phases interleaved
+ uint8x8x2_t o;
+ o.val[0] = vqrshrun_n_s16(even, 4);
+ o.val[1] = vqrshrun_n_s16(odd, 4);
+ vst2_u8(out + i*2, o);
+#endif
+
+ // "previous" value for next iter
+ t1 = 3*in_near[i+7] + in_far[i+7];
+ }
+
+ t0 = t1;
+ t1 = 3*in_near[i] + in_far[i];
+ out[i*2] = stbi__div16(3*t1 + t0 + 8);
+
+ for (++i; i < w; ++i) {
+ t0 = t1;
+ t1 = 3*in_near[i]+in_far[i];
+ out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+ out[i*2 ] = stbi__div16(3*t1 + t0 + 8);
+ }
+ out[w*2-1] = stbi__div4(t1+2);
+
+ STBI_NOTUSED(hs);
+
+ return out;
+}
+#endif
+
+static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+ // resample with nearest-neighbor
+ int i,j;
+ STBI_NOTUSED(in_far);
+ for (i=0; i < w; ++i)
+ for (j=0; j < hs; ++j)
+ out[i*hs+j] = in_near[i];
+ return out;
+}
+
+// this is a reduced-precision calculation of YCbCr-to-RGB introduced
+// to make sure the code produces the same results in both SIMD and scalar
+#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8)
+static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)
+{
+ int i;
+ for (i=0; i < count; ++i) {
+ int y_fixed = (y[i] << 20) + (1<<19); // rounding
+ int r,g,b;
+ int cr = pcr[i] - 128;
+ int cb = pcb[i] - 128;
+ r = y_fixed + cr* stbi__float2fixed(1.40200f);
+ g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+ b = y_fixed + cb* stbi__float2fixed(1.77200f);
+ r >>= 20;
+ g >>= 20;
+ b >>= 20;
+ if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+ if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+ if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+ out[0] = (stbi_uc)r;
+ out[1] = (stbi_uc)g;
+ out[2] = (stbi_uc)b;
+ out[3] = 255;
+ out += step;
+ }
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step)
+{
+ int i = 0;
+
+#ifdef STBI_SSE2
+ // step == 3 is pretty ugly on the final interleave, and i'm not convinced
+ // it's useful in practice (you wouldn't use it for textures, for example).
+ // so just accelerate step == 4 case.
+ if (step == 4) {
+ // this is a fairly straightforward implementation and not super-optimized.
+ __m128i signflip = _mm_set1_epi8(-0x80);
+ __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f));
+ __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f));
+ __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f));
+ __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f));
+ __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128);
+ __m128i xw = _mm_set1_epi16(255); // alpha channel
+
+ for (; i+7 < count; i += 8) {
+ // load
+ __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i));
+ __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i));
+ __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i));
+ __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128
+ __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128
+
+ // unpack to short (and left-shift cr, cb by 8)
+ __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes);
+ __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased);
+ __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased);
+
+ // color transform
+ __m128i yws = _mm_srli_epi16(yw, 4);
+ __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw);
+ __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw);
+ __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1);
+ __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1);
+ __m128i rws = _mm_add_epi16(cr0, yws);
+ __m128i gwt = _mm_add_epi16(cb0, yws);
+ __m128i bws = _mm_add_epi16(yws, cb1);
+ __m128i gws = _mm_add_epi16(gwt, cr1);
+
+ // descale
+ __m128i rw = _mm_srai_epi16(rws, 4);
+ __m128i bw = _mm_srai_epi16(bws, 4);
+ __m128i gw = _mm_srai_epi16(gws, 4);
+
+ // back to byte, set up for transpose
+ __m128i brb = _mm_packus_epi16(rw, bw);
+ __m128i gxb = _mm_packus_epi16(gw, xw);
+
+ // transpose to interleave channels
+ __m128i t0 = _mm_unpacklo_epi8(brb, gxb);
+ __m128i t1 = _mm_unpackhi_epi8(brb, gxb);
+ __m128i o0 = _mm_unpacklo_epi16(t0, t1);
+ __m128i o1 = _mm_unpackhi_epi16(t0, t1);
+
+ // store
+ _mm_storeu_si128((__m128i *) (out + 0), o0);
+ _mm_storeu_si128((__m128i *) (out + 16), o1);
+ out += 32;
+ }
+ }
+#endif
+
+#ifdef STBI_NEON
+ // in this version, step=3 support would be easy to add. but is there demand?
+ if (step == 4) {
+ // this is a fairly straightforward implementation and not super-optimized.
+ uint8x8_t signflip = vdup_n_u8(0x80);
+ int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f));
+ int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f));
+ int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f));
+ int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f));
+
+ for (; i+7 < count; i += 8) {
+ // load
+ uint8x8_t y_bytes = vld1_u8(y + i);
+ uint8x8_t cr_bytes = vld1_u8(pcr + i);
+ uint8x8_t cb_bytes = vld1_u8(pcb + i);
+ int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip));
+ int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip));
+
+ // expand to s16
+ int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4));
+ int16x8_t crw = vshll_n_s8(cr_biased, 7);
+ int16x8_t cbw = vshll_n_s8(cb_biased, 7);
+
+ // color transform
+ int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0);
+ int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0);
+ int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1);
+ int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1);
+ int16x8_t rws = vaddq_s16(yws, cr0);
+ int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1);
+ int16x8_t bws = vaddq_s16(yws, cb1);
+
+ // undo scaling, round, convert to byte
+ uint8x8x4_t o;
+ o.val[0] = vqrshrun_n_s16(rws, 4);
+ o.val[1] = vqrshrun_n_s16(gws, 4);
+ o.val[2] = vqrshrun_n_s16(bws, 4);
+ o.val[3] = vdup_n_u8(255);
+
+ // store, interleaving r/g/b/a
+ vst4_u8(out, o);
+ out += 8*4;
+ }
+ }
+#endif
+
+ for (; i < count; ++i) {
+ int y_fixed = (y[i] << 20) + (1<<19); // rounding
+ int r,g,b;
+ int cr = pcr[i] - 128;
+ int cb = pcb[i] - 128;
+ r = y_fixed + cr* stbi__float2fixed(1.40200f);
+ g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+ b = y_fixed + cb* stbi__float2fixed(1.77200f);
+ r >>= 20;
+ g >>= 20;
+ b >>= 20;
+ if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+ if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+ if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+ out[0] = (stbi_uc)r;
+ out[1] = (stbi_uc)g;
+ out[2] = (stbi_uc)b;
+ out[3] = 255;
+ out += step;
+ }
+}
+#endif
+
+// set up the kernels
+static void stbi__setup_jpeg(stbi__jpeg *j)
+{
+ j->idct_block_kernel = stbi__idct_block;
+ j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row;
+ j->resample_row_hv_2_kernel = stbi__resample_row_hv_2;
+
+#ifdef STBI_SSE2
+ if (stbi__sse2_available()) {
+ j->idct_block_kernel = stbi__idct_simd;
+ j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+ j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+ }
+#endif
+
+#ifdef STBI_NEON
+ j->idct_block_kernel = stbi__idct_simd;
+ j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+ j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+#endif
+}
+
+// clean up the temporary component buffers
+static void stbi__cleanup_jpeg(stbi__jpeg *j)
+{
+ stbi__free_jpeg_components(j, j->s->img_n, 0);
+}
+
+typedef struct
+{
+ resample_row_func resample;
+ stbi_uc *line0,*line1;
+ int hs,vs; // expansion factor in each axis
+ int w_lores; // horizontal pixels pre-expansion
+ int ystep; // how far through vertical expansion we are
+ int ypos; // which pre-expansion row we're on
+} stbi__resample;
+
+// fast 0..255 * 0..255 => 0..255 rounded multiplication
+static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y)
+{
+ unsigned int t = x*y + 128;
+ return (stbi_uc) ((t + (t >>8)) >> 8);
+}
+
+static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp)
+{
+ int n, decode_n, is_rgb;
+ z->s->img_n = 0; // make stbi__cleanup_jpeg safe
+
+ // validate req_comp
+ if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+
+ // load a jpeg image from whichever source, but leave in YCbCr format
+ if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; }
+
+ // determine actual number of components to generate
+ n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1;
+
+ is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif));
+
+ if (z->s->img_n == 3 && n < 3 && !is_rgb)
+ decode_n = 1;
+ else
+ decode_n = z->s->img_n;
+
+ // resample and color-convert
+ {
+ int k;
+ unsigned int i,j;
+ stbi_uc *output;
+ stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL };
+
+ stbi__resample res_comp[4];
+
+ for (k=0; k < decode_n; ++k) {
+ stbi__resample *r = &res_comp[k];
+
+ // allocate line buffer big enough for upsampling off the edges
+ // with upsample factor of 4
+ z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3);
+ if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+ r->hs = z->img_h_max / z->img_comp[k].h;
+ r->vs = z->img_v_max / z->img_comp[k].v;
+ r->ystep = r->vs >> 1;
+ r->w_lores = (z->s->img_x + r->hs-1) / r->hs;
+ r->ypos = 0;
+ r->line0 = r->line1 = z->img_comp[k].data;
+
+ if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1;
+ else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2;
+ else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2;
+ else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel;
+ else r->resample = stbi__resample_row_generic;
+ }
+
+ // can't error after this so, this is safe
+ output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1);
+ if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+ // now go ahead and resample
+ for (j=0; j < z->s->img_y; ++j) {
+ stbi_uc *out = output + n * z->s->img_x * j;
+ for (k=0; k < decode_n; ++k) {
+ stbi__resample *r = &res_comp[k];
+ int y_bot = r->ystep >= (r->vs >> 1);
+ coutput[k] = r->resample(z->img_comp[k].linebuf,
+ y_bot ? r->line1 : r->line0,
+ y_bot ? r->line0 : r->line1,
+ r->w_lores, r->hs);
+ if (++r->ystep >= r->vs) {
+ r->ystep = 0;
+ r->line0 = r->line1;
+ if (++r->ypos < z->img_comp[k].y)
+ r->line1 += z->img_comp[k].w2;
+ }
+ }
+ if (n >= 3) {
+ stbi_uc *y = coutput[0];
+ if (z->s->img_n == 3) {
+ if (is_rgb) {
+ for (i=0; i < z->s->img_x; ++i) {
+ out[0] = y[i];
+ out[1] = coutput[1][i];
+ out[2] = coutput[2][i];
+ out[3] = 255;
+ out += n;
+ }
+ } else {
+ z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+ }
+ } else if (z->s->img_n == 4) {
+ if (z->app14_color_transform == 0) { // CMYK
+ for (i=0; i < z->s->img_x; ++i) {
+ stbi_uc m = coutput[3][i];
+ out[0] = stbi__blinn_8x8(coutput[0][i], m);
+ out[1] = stbi__blinn_8x8(coutput[1][i], m);
+ out[2] = stbi__blinn_8x8(coutput[2][i], m);
+ out[3] = 255;
+ out += n;
+ }
+ } else if (z->app14_color_transform == 2) { // YCCK
+ z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+ for (i=0; i < z->s->img_x; ++i) {
+ stbi_uc m = coutput[3][i];
+ out[0] = stbi__blinn_8x8(255 - out[0], m);
+ out[1] = stbi__blinn_8x8(255 - out[1], m);
+ out[2] = stbi__blinn_8x8(255 - out[2], m);
+ out += n;
+ }
+ } else { // YCbCr + alpha? Ignore the fourth channel for now
+ z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+ }
+ } else
+ for (i=0; i < z->s->img_x; ++i) {
+ out[0] = out[1] = out[2] = y[i];
+ out[3] = 255; // not used if n==3
+ out += n;
+ }
+ } else {
+ if (is_rgb) {
+ if (n == 1)
+ for (i=0; i < z->s->img_x; ++i)
+ *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+ else {
+ for (i=0; i < z->s->img_x; ++i, out += 2) {
+ out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+ out[1] = 255;
+ }
+ }
+ } else if (z->s->img_n == 4 && z->app14_color_transform == 0) {
+ for (i=0; i < z->s->img_x; ++i) {
+ stbi_uc m = coutput[3][i];
+ stbi_uc r = stbi__blinn_8x8(coutput[0][i], m);
+ stbi_uc g = stbi__blinn_8x8(coutput[1][i], m);
+ stbi_uc b = stbi__blinn_8x8(coutput[2][i], m);
+ out[0] = stbi__compute_y(r, g, b);
+ out[1] = 255;
+ out += n;
+ }
+ } else if (z->s->img_n == 4 && z->app14_color_transform == 2) {
+ for (i=0; i < z->s->img_x; ++i) {
+ out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]);
+ out[1] = 255;
+ out += n;
+ }
+ } else {
+ stbi_uc *y = coutput[0];
+ if (n == 1)
+ for (i=0; i < z->s->img_x; ++i) out[i] = y[i];
+ else
+ for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; }
+ }
+ }
+ }
+ stbi__cleanup_jpeg(z);
+ *out_x = z->s->img_x;
+ *out_y = z->s->img_y;
+ if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output
+ return output;
+ }
+}
+
+static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ unsigned char* result;
+ stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg));
+ STBI_NOTUSED(ri);
+ j->s = s;
+ stbi__setup_jpeg(j);
+ result = load_jpeg_image(j, x,y,comp,req_comp);
+ STBI_FREE(j);
+ return result;
+}
+
+static int stbi__jpeg_test(stbi__context *s)
+{
+ int r;
+ stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg));
+ j->s = s;
+ stbi__setup_jpeg(j);
+ r = stbi__decode_jpeg_header(j, STBI__SCAN_type);
+ stbi__rewind(s);
+ STBI_FREE(j);
+ return r;
+}
+
+static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp)
+{
+ if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) {
+ stbi__rewind( j->s );
+ return 0;
+ }
+ if (x) *x = j->s->img_x;
+ if (y) *y = j->s->img_y;
+ if (comp) *comp = j->s->img_n >= 3 ? 3 : 1;
+ return 1;
+}
+
+static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ int result;
+ stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg)));
+ j->s = s;
+ result = stbi__jpeg_info_raw(j, x, y, comp);
+ STBI_FREE(j);
+ return result;
+}
+#endif
+
+// public domain zlib decode v0.2 Sean Barrett 2006-11-18
+// simple implementation
+// - all input must be provided in an upfront buffer
+// - all output is written to a single output buffer (can malloc/realloc)
+// performance
+// - fast huffman
+
+#ifndef STBI_NO_ZLIB
+
+// fast-way is faster to check than jpeg huffman, but slow way is slower
+#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables
+#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1)
+
+// zlib-style huffman encoding
+// (jpegs packs from left, zlib from right, so can't share code)
+typedef struct
+{
+ stbi__uint16 fast[1 << STBI__ZFAST_BITS];
+ stbi__uint16 firstcode[16];
+ int maxcode[17];
+ stbi__uint16 firstsymbol[16];
+ stbi_uc size[288];
+ stbi__uint16 value[288];
+} stbi__zhuffman;
+
+stbi_inline static int stbi__bitreverse16(int n)
+{
+ n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1);
+ n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2);
+ n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4);
+ n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8);
+ return n;
+}
+
+stbi_inline static int stbi__bit_reverse(int v, int bits)
+{
+ STBI_ASSERT(bits <= 16);
+ // to bit reverse n bits, reverse 16 and shift
+ // e.g. 11 bits, bit reverse and shift away 5
+ return stbi__bitreverse16(v) >> (16-bits);
+}
+
+static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num)
+{
+ int i,k=0;
+ int code, next_code[16], sizes[17];
+
+ // DEFLATE spec for generating codes
+ memset(sizes, 0, sizeof(sizes));
+ memset(z->fast, 0, sizeof(z->fast));
+ for (i=0; i < num; ++i)
+ ++sizes[sizelist[i]];
+ sizes[0] = 0;
+ for (i=1; i < 16; ++i)
+ if (sizes[i] > (1 << i))
+ return stbi__err("bad sizes", "Corrupt PNG");
+ code = 0;
+ for (i=1; i < 16; ++i) {
+ next_code[i] = code;
+ z->firstcode[i] = (stbi__uint16) code;
+ z->firstsymbol[i] = (stbi__uint16) k;
+ code = (code + sizes[i]);
+ if (sizes[i])
+ if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG");
+ z->maxcode[i] = code << (16-i); // preshift for inner loop
+ code <<= 1;
+ k += sizes[i];
+ }
+ z->maxcode[16] = 0x10000; // sentinel
+ for (i=0; i < num; ++i) {
+ int s = sizelist[i];
+ if (s) {
+ int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];
+ stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i);
+ z->size [c] = (stbi_uc ) s;
+ z->value[c] = (stbi__uint16) i;
+ if (s <= STBI__ZFAST_BITS) {
+ int j = stbi__bit_reverse(next_code[s],s);
+ while (j < (1 << STBI__ZFAST_BITS)) {
+ z->fast[j] = fastv;
+ j += (1 << s);
+ }
+ }
+ ++next_code[s];
+ }
+ }
+ return 1;
+}
+
+// zlib-from-memory implementation for PNG reading
+// because PNG allows splitting the zlib stream arbitrarily,
+// and it's annoying structurally to have PNG call ZLIB call PNG,
+// we require PNG read all the IDATs and combine them into a single
+// memory buffer
+
+typedef struct
+{
+ stbi_uc *zbuffer, *zbuffer_end;
+ int num_bits;
+ stbi__uint32 code_buffer;
+
+ char *zout;
+ char *zout_start;
+ char *zout_end;
+ int z_expandable;
+
+ stbi__zhuffman z_length, z_distance;
+} stbi__zbuf;
+
+stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z)
+{
+ if (z->zbuffer >= z->zbuffer_end) return 0;
+ return *z->zbuffer++;
+}
+
+static void stbi__fill_bits(stbi__zbuf *z)
+{
+ do {
+ STBI_ASSERT(z->code_buffer < (1U << z->num_bits));
+ z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits;
+ z->num_bits += 8;
+ } while (z->num_bits <= 24);
+}
+
+stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n)
+{
+ unsigned int k;
+ if (z->num_bits < n) stbi__fill_bits(z);
+ k = z->code_buffer & ((1 << n) - 1);
+ z->code_buffer >>= n;
+ z->num_bits -= n;
+ return k;
+}
+
+static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)
+{
+ int b,s,k;
+ // not resolved by fast table, so compute it the slow way
+ // use jpeg approach, which requires MSbits at top
+ k = stbi__bit_reverse(a->code_buffer, 16);
+ for (s=STBI__ZFAST_BITS+1; ; ++s)
+ if (k < z->maxcode[s])
+ break;
+ if (s == 16) return -1; // invalid code!
+ // code size is s, so:
+ b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];
+ STBI_ASSERT(z->size[b] == s);
+ a->code_buffer >>= s;
+ a->num_bits -= s;
+ return z->value[b];
+}
+
+stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)
+{
+ int b,s;
+ if (a->num_bits < 16) stbi__fill_bits(a);
+ b = z->fast[a->code_buffer & STBI__ZFAST_MASK];
+ if (b) {
+ s = b >> 9;
+ a->code_buffer >>= s;
+ a->num_bits -= s;
+ return b & 511;
+ }
+ return stbi__zhuffman_decode_slowpath(a, z);
+}
+
+static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes
+{
+ char *q;
+ int cur, limit, old_limit;
+ z->zout = zout;
+ if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG");
+ cur = (int) (z->zout - z->zout_start);
+ limit = old_limit = (int) (z->zout_end - z->zout_start);
+ while (cur + n > limit)
+ limit *= 2;
+ q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit);
+ STBI_NOTUSED(old_limit);
+ if (q == NULL) return stbi__err("outofmem", "Out of memory");
+ z->zout_start = q;
+ z->zout = q + cur;
+ z->zout_end = q + limit;
+ return 1;
+}
+
+static const int stbi__zlength_base[31] = {
+ 3,4,5,6,7,8,9,10,11,13,
+ 15,17,19,23,27,31,35,43,51,59,
+ 67,83,99,115,131,163,195,227,258,0,0 };
+
+static const int stbi__zlength_extra[31]=
+{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };
+
+static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,
+257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};
+
+static const int stbi__zdist_extra[32] =
+{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+static int stbi__parse_huffman_block(stbi__zbuf *a)
+{
+ char *zout = a->zout;
+ for(;;) {
+ int z = stbi__zhuffman_decode(a, &a->z_length);
+ if (z < 256) {
+ if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes
+ if (zout >= a->zout_end) {
+ if (!stbi__zexpand(a, zout, 1)) return 0;
+ zout = a->zout;
+ }
+ *zout++ = (char) z;
+ } else {
+ stbi_uc *p;
+ int len,dist;
+ if (z == 256) {
+ a->zout = zout;
+ return 1;
+ }
+ z -= 257;
+ len = stbi__zlength_base[z];
+ if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);
+ z = stbi__zhuffman_decode(a, &a->z_distance);
+ if (z < 0) return stbi__err("bad huffman code","Corrupt PNG");
+ dist = stbi__zdist_base[z];
+ if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);
+ if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG");
+ if (zout + len > a->zout_end) {
+ if (!stbi__zexpand(a, zout, len)) return 0;
+ zout = a->zout;
+ }
+ p = (stbi_uc *) (zout - dist);
+ if (dist == 1) { // run of one byte; common in images.
+ stbi_uc v = *p;
+ if (len) { do *zout++ = v; while (--len); }
+ } else {
+ if (len) { do *zout++ = *p++; while (--len); }
+ }
+ }
+ }
+}
+
+static int stbi__compute_huffman_codes(stbi__zbuf *a)
+{
+ static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };
+ stbi__zhuffman z_codelength;
+ stbi_uc lencodes[286+32+137];//padding for maximum single op
+ stbi_uc codelength_sizes[19];
+ int i,n;
+
+ int hlit = stbi__zreceive(a,5) + 257;
+ int hdist = stbi__zreceive(a,5) + 1;
+ int hclen = stbi__zreceive(a,4) + 4;
+ int ntot = hlit + hdist;
+
+ memset(codelength_sizes, 0, sizeof(codelength_sizes));
+ for (i=0; i < hclen; ++i) {
+ int s = stbi__zreceive(a,3);
+ codelength_sizes[length_dezigzag[i]] = (stbi_uc) s;
+ }
+ if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;
+
+ n = 0;
+ while (n < ntot) {
+ int c = stbi__zhuffman_decode(a, &z_codelength);
+ if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG");
+ if (c < 16)
+ lencodes[n++] = (stbi_uc) c;
+ else {
+ stbi_uc fill = 0;
+ if (c == 16) {
+ c = stbi__zreceive(a,2)+3;
+ if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG");
+ fill = lencodes[n-1];
+ } else if (c == 17)
+ c = stbi__zreceive(a,3)+3;
+ else {
+ STBI_ASSERT(c == 18);
+ c = stbi__zreceive(a,7)+11;
+ }
+ if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG");
+ memset(lencodes+n, fill, c);
+ n += c;
+ }
+ }
+ if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG");
+ if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;
+ if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;
+ return 1;
+}
+
+static int stbi__parse_uncompressed_block(stbi__zbuf *a)
+{
+ stbi_uc header[4];
+ int len,nlen,k;
+ if (a->num_bits & 7)
+ stbi__zreceive(a, a->num_bits & 7); // discard
+ // drain the bit-packed data into header
+ k = 0;
+ while (a->num_bits > 0) {
+ header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check
+ a->code_buffer >>= 8;
+ a->num_bits -= 8;
+ }
+ STBI_ASSERT(a->num_bits == 0);
+ // now fill header the normal way
+ while (k < 4)
+ header[k++] = stbi__zget8(a);
+ len = header[1] * 256 + header[0];
+ nlen = header[3] * 256 + header[2];
+ if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG");
+ if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG");
+ if (a->zout + len > a->zout_end)
+ if (!stbi__zexpand(a, a->zout, len)) return 0;
+ memcpy(a->zout, a->zbuffer, len);
+ a->zbuffer += len;
+ a->zout += len;
+ return 1;
+}
+
+static int stbi__parse_zlib_header(stbi__zbuf *a)
+{
+ int cmf = stbi__zget8(a);
+ int cm = cmf & 15;
+ /* int cinfo = cmf >> 4; */
+ int flg = stbi__zget8(a);
+ if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
+ if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png
+ if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png
+ // window = 1 << (8 + cinfo)... but who cares, we fully buffer output
+ return 1;
+}
+
+static const stbi_uc stbi__zdefault_length[288] =
+{
+ 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+ 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+ 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+ 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+ 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+ 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+ 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+ 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+ 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8
+};
+static const stbi_uc stbi__zdefault_distance[32] =
+{
+ 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5
+};
+/*
+Init algorithm:
+{
+ int i; // use <= to match clearly with spec
+ for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8;
+ for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9;
+ for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7;
+ for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8;
+
+ for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5;
+}
+*/
+
+static int stbi__parse_zlib(stbi__zbuf *a, int parse_header)
+{
+ int final, type;
+ if (parse_header)
+ if (!stbi__parse_zlib_header(a)) return 0;
+ a->num_bits = 0;
+ a->code_buffer = 0;
+ do {
+ final = stbi__zreceive(a,1);
+ type = stbi__zreceive(a,2);
+ if (type == 0) {
+ if (!stbi__parse_uncompressed_block(a)) return 0;
+ } else if (type == 3) {
+ return 0;
+ } else {
+ if (type == 1) {
+ // use fixed code lengths
+ if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , 288)) return 0;
+ if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0;
+ } else {
+ if (!stbi__compute_huffman_codes(a)) return 0;
+ }
+ if (!stbi__parse_huffman_block(a)) return 0;
+ }
+ } while (!final);
+ return 1;
+}
+
+static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)
+{
+ a->zout_start = obuf;
+ a->zout = obuf;
+ a->zout_end = obuf + olen;
+ a->z_expandable = exp;
+
+ return stbi__parse_zlib(a, parse_header);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)
+{
+ stbi__zbuf a;
+ char *p = (char *) stbi__malloc(initial_size);
+ if (p == NULL) return NULL;
+ a.zbuffer = (stbi_uc *) buffer;
+ a.zbuffer_end = (stbi_uc *) buffer + len;
+ if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {
+ if (outlen) *outlen = (int) (a.zout - a.zout_start);
+ return a.zout_start;
+ } else {
+ STBI_FREE(a.zout_start);
+ return NULL;
+ }
+}
+
+STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)
+{
+ return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)
+{
+ stbi__zbuf a;
+ char *p = (char *) stbi__malloc(initial_size);
+ if (p == NULL) return NULL;
+ a.zbuffer = (stbi_uc *) buffer;
+ a.zbuffer_end = (stbi_uc *) buffer + len;
+ if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {
+ if (outlen) *outlen = (int) (a.zout - a.zout_start);
+ return a.zout_start;
+ } else {
+ STBI_FREE(a.zout_start);
+ return NULL;
+ }
+}
+
+STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)
+{
+ stbi__zbuf a;
+ a.zbuffer = (stbi_uc *) ibuffer;
+ a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+ if (stbi__do_zlib(&a, obuffer, olen, 0, 1))
+ return (int) (a.zout - a.zout_start);
+ else
+ return -1;
+}
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)
+{
+ stbi__zbuf a;
+ char *p = (char *) stbi__malloc(16384);
+ if (p == NULL) return NULL;
+ a.zbuffer = (stbi_uc *) buffer;
+ a.zbuffer_end = (stbi_uc *) buffer+len;
+ if (stbi__do_zlib(&a, p, 16384, 1, 0)) {
+ if (outlen) *outlen = (int) (a.zout - a.zout_start);
+ return a.zout_start;
+ } else {
+ STBI_FREE(a.zout_start);
+ return NULL;
+ }
+}
+
+STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)
+{
+ stbi__zbuf a;
+ a.zbuffer = (stbi_uc *) ibuffer;
+ a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+ if (stbi__do_zlib(&a, obuffer, olen, 0, 0))
+ return (int) (a.zout - a.zout_start);
+ else
+ return -1;
+}
+#endif
+
+// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18
+// simple implementation
+// - only 8-bit samples
+// - no CRC checking
+// - allocates lots of intermediate memory
+// - avoids problem of streaming data between subsystems
+// - avoids explicit window management
+// performance
+// - uses stb_zlib, a PD zlib implementation with fast huffman decoding
+
+#ifndef STBI_NO_PNG
+typedef struct
+{
+ stbi__uint32 length;
+ stbi__uint32 type;
+} stbi__pngchunk;
+
+static stbi__pngchunk stbi__get_chunk_header(stbi__context *s)
+{
+ stbi__pngchunk c;
+ c.length = stbi__get32be(s);
+ c.type = stbi__get32be(s);
+ return c;
+}
+
+static int stbi__check_png_header(stbi__context *s)
+{
+ static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 };
+ int i;
+ for (i=0; i < 8; ++i)
+ if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG");
+ return 1;
+}
+
+typedef struct
+{
+ stbi__context *s;
+ stbi_uc *idata, *expanded, *out;
+ int depth;
+} stbi__png;
+
+
+enum {
+ STBI__F_none=0,
+ STBI__F_sub=1,
+ STBI__F_up=2,
+ STBI__F_avg=3,
+ STBI__F_paeth=4,
+ // synthetic filters used for first scanline to avoid needing a dummy row of 0s
+ STBI__F_avg_first,
+ STBI__F_paeth_first
+};
+
+static stbi_uc first_row_filter[5] =
+{
+ STBI__F_none,
+ STBI__F_sub,
+ STBI__F_none,
+ STBI__F_avg_first,
+ STBI__F_paeth_first
+};
+
+static int stbi__paeth(int a, int b, int c)
+{
+ int p = a + b - c;
+ int pa = abs(p-a);
+ int pb = abs(p-b);
+ int pc = abs(p-c);
+ if (pa <= pb && pa <= pc) return a;
+ if (pb <= pc) return b;
+ return c;
+}
+
+static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 };
+
+// create the png data from post-deflated data
+static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color)
+{
+ int bytes = (depth == 16? 2 : 1);
+ stbi__context *s = a->s;
+ stbi__uint32 i,j,stride = x*out_n*bytes;
+ stbi__uint32 img_len, img_width_bytes;
+ int k;
+ int img_n = s->img_n; // copy it into a local for later
+
+ int output_bytes = out_n*bytes;
+ int filter_bytes = img_n*bytes;
+ int width = x;
+
+ STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1);
+ a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into
+ if (!a->out) return stbi__err("outofmem", "Out of memory");
+
+ if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG");
+ img_width_bytes = (((img_n * x * depth) + 7) >> 3);
+ img_len = (img_width_bytes + 1) * y;
+
+ // we used to check for exact match between raw_len and img_len on non-interlaced PNGs,
+ // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),
+ // so just check for raw_len < img_len always.
+ if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
+
+ for (j=0; j < y; ++j) {
+ stbi_uc *cur = a->out + stride*j;
+ stbi_uc *prior;
+ int filter = *raw++;
+
+ if (filter > 4)
+ return stbi__err("invalid filter","Corrupt PNG");
+
+ if (depth < 8) {
+ STBI_ASSERT(img_width_bytes <= x);
+ cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place
+ filter_bytes = 1;
+ width = img_width_bytes;
+ }
+ prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above
+
+ // if first row, use special filter that doesn't sample previous row
+ if (j == 0) filter = first_row_filter[filter];
+
+ // handle first byte explicitly
+ for (k=0; k < filter_bytes; ++k) {
+ switch (filter) {
+ case STBI__F_none : cur[k] = raw[k]; break;
+ case STBI__F_sub : cur[k] = raw[k]; break;
+ case STBI__F_up : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;
+ case STBI__F_avg : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break;
+ case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;
+ case STBI__F_avg_first : cur[k] = raw[k]; break;
+ case STBI__F_paeth_first: cur[k] = raw[k]; break;
+ }
+ }
+
+ if (depth == 8) {
+ if (img_n != out_n)
+ cur[img_n] = 255; // first pixel
+ raw += img_n;
+ cur += out_n;
+ prior += out_n;
+ } else if (depth == 16) {
+ if (img_n != out_n) {
+ cur[filter_bytes] = 255; // first pixel top byte
+ cur[filter_bytes+1] = 255; // first pixel bottom byte
+ }
+ raw += filter_bytes;
+ cur += output_bytes;
+ prior += output_bytes;
+ } else {
+ raw += 1;
+ cur += 1;
+ prior += 1;
+ }
+
+ // this is a little gross, so that we don't switch per-pixel or per-component
+ if (depth < 8 || img_n == out_n) {
+ int nk = (width - 1)*filter_bytes;
+ #define STBI__CASE(f) \
+ case f: \
+ for (k=0; k < nk; ++k)
+ switch (filter) {
+ // "none" filter turns into a memcpy here; make that explicit.
+ case STBI__F_none: memcpy(cur, raw, nk); break;
+ STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break;
+ STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+ STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break;
+ STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break;
+ STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break;
+ STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break;
+ }
+ #undef STBI__CASE
+ raw += nk;
+ } else {
+ STBI_ASSERT(img_n+1 == out_n);
+ #define STBI__CASE(f) \
+ case f: \
+ for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \
+ for (k=0; k < filter_bytes; ++k)
+ switch (filter) {
+ STBI__CASE(STBI__F_none) { cur[k] = raw[k]; } break;
+ STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break;
+ STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+ STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break;
+ STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break;
+ STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break;
+ STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break;
+ }
+ #undef STBI__CASE
+
+ // the loop above sets the high byte of the pixels' alpha, but for
+ // 16 bit png files we also need the low byte set. we'll do that here.
+ if (depth == 16) {
+ cur = a->out + stride*j; // start at the beginning of the row again
+ for (i=0; i < x; ++i,cur+=output_bytes) {
+ cur[filter_bytes+1] = 255;
+ }
+ }
+ }
+ }
+
+ // we make a separate pass to expand bits to pixels; for performance,
+ // this could run two scanlines behind the above code, so it won't
+ // intefere with filtering but will still be in the cache.
+ if (depth < 8) {
+ for (j=0; j < y; ++j) {
+ stbi_uc *cur = a->out + stride*j;
+ stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes;
+ // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit
+ // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop
+ stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range
+
+ // note that the final byte might overshoot and write more data than desired.
+ // we can allocate enough data that this never writes out of memory, but it
+ // could also overwrite the next scanline. can it overwrite non-empty data
+ // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel.
+ // so we need to explicitly clamp the final ones
+
+ if (depth == 4) {
+ for (k=x*img_n; k >= 2; k-=2, ++in) {
+ *cur++ = scale * ((*in >> 4) );
+ *cur++ = scale * ((*in ) & 0x0f);
+ }
+ if (k > 0) *cur++ = scale * ((*in >> 4) );
+ } else if (depth == 2) {
+ for (k=x*img_n; k >= 4; k-=4, ++in) {
+ *cur++ = scale * ((*in >> 6) );
+ *cur++ = scale * ((*in >> 4) & 0x03);
+ *cur++ = scale * ((*in >> 2) & 0x03);
+ *cur++ = scale * ((*in ) & 0x03);
+ }
+ if (k > 0) *cur++ = scale * ((*in >> 6) );
+ if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03);
+ if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03);
+ } else if (depth == 1) {
+ for (k=x*img_n; k >= 8; k-=8, ++in) {
+ *cur++ = scale * ((*in >> 7) );
+ *cur++ = scale * ((*in >> 6) & 0x01);
+ *cur++ = scale * ((*in >> 5) & 0x01);
+ *cur++ = scale * ((*in >> 4) & 0x01);
+ *cur++ = scale * ((*in >> 3) & 0x01);
+ *cur++ = scale * ((*in >> 2) & 0x01);
+ *cur++ = scale * ((*in >> 1) & 0x01);
+ *cur++ = scale * ((*in ) & 0x01);
+ }
+ if (k > 0) *cur++ = scale * ((*in >> 7) );
+ if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01);
+ if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01);
+ if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01);
+ if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01);
+ if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01);
+ if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01);
+ }
+ if (img_n != out_n) {
+ int q;
+ // insert alpha = 255
+ cur = a->out + stride*j;
+ if (img_n == 1) {
+ for (q=x-1; q >= 0; --q) {
+ cur[q*2+1] = 255;
+ cur[q*2+0] = cur[q];
+ }
+ } else {
+ STBI_ASSERT(img_n == 3);
+ for (q=x-1; q >= 0; --q) {
+ cur[q*4+3] = 255;
+ cur[q*4+2] = cur[q*3+2];
+ cur[q*4+1] = cur[q*3+1];
+ cur[q*4+0] = cur[q*3+0];
+ }
+ }
+ }
+ }
+ } else if (depth == 16) {
+ // force the image data from big-endian to platform-native.
+ // this is done in a separate pass due to the decoding relying
+ // on the data being untouched, but could probably be done
+ // per-line during decode if care is taken.
+ stbi_uc *cur = a->out;
+ stbi__uint16 *cur16 = (stbi__uint16*)cur;
+
+ for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) {
+ *cur16 = (cur[0] << 8) | cur[1];
+ }
+ }
+
+ return 1;
+}
+
+static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced)
+{
+ int bytes = (depth == 16 ? 2 : 1);
+ int out_bytes = out_n * bytes;
+ stbi_uc *final;
+ int p;
+ if (!interlaced)
+ return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);
+
+ // de-interlacing
+ final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0);
+ for (p=0; p < 7; ++p) {
+ int xorig[] = { 0,4,0,2,0,1,0 };
+ int yorig[] = { 0,0,4,0,2,0,1 };
+ int xspc[] = { 8,8,4,4,2,2,1 };
+ int yspc[] = { 8,8,8,4,4,2,2 };
+ int i,j,x,y;
+ // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1
+ x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];
+ y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];
+ if (x && y) {
+ stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;
+ if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {
+ STBI_FREE(final);
+ return 0;
+ }
+ for (j=0; j < y; ++j) {
+ for (i=0; i < x; ++i) {
+ int out_y = j*yspc[p]+yorig[p];
+ int out_x = i*xspc[p]+xorig[p];
+ memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes,
+ a->out + (j*x+i)*out_bytes, out_bytes);
+ }
+ }
+ STBI_FREE(a->out);
+ image_data += img_len;
+ image_data_len -= img_len;
+ }
+ }
+ a->out = final;
+
+ return 1;
+}
+
+static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n)
+{
+ stbi__context *s = z->s;
+ stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+ stbi_uc *p = z->out;
+
+ // compute color-based transparency, assuming we've
+ // already got 255 as the alpha value in the output
+ STBI_ASSERT(out_n == 2 || out_n == 4);
+
+ if (out_n == 2) {
+ for (i=0; i < pixel_count; ++i) {
+ p[1] = (p[0] == tc[0] ? 0 : 255);
+ p += 2;
+ }
+ } else {
+ for (i=0; i < pixel_count; ++i) {
+ if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+ p[3] = 0;
+ p += 4;
+ }
+ }
+ return 1;
+}
+
+static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n)
+{
+ stbi__context *s = z->s;
+ stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+ stbi__uint16 *p = (stbi__uint16*) z->out;
+
+ // compute color-based transparency, assuming we've
+ // already got 65535 as the alpha value in the output
+ STBI_ASSERT(out_n == 2 || out_n == 4);
+
+ if (out_n == 2) {
+ for (i = 0; i < pixel_count; ++i) {
+ p[1] = (p[0] == tc[0] ? 0 : 65535);
+ p += 2;
+ }
+ } else {
+ for (i = 0; i < pixel_count; ++i) {
+ if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+ p[3] = 0;
+ p += 4;
+ }
+ }
+ return 1;
+}
+
+static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n)
+{
+ stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y;
+ stbi_uc *p, *temp_out, *orig = a->out;
+
+ p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0);
+ if (p == NULL) return stbi__err("outofmem", "Out of memory");
+
+ // between here and free(out) below, exitting would leak
+ temp_out = p;
+
+ if (pal_img_n == 3) {
+ for (i=0; i < pixel_count; ++i) {
+ int n = orig[i]*4;
+ p[0] = palette[n ];
+ p[1] = palette[n+1];
+ p[2] = palette[n+2];
+ p += 3;
+ }
+ } else {
+ for (i=0; i < pixel_count; ++i) {
+ int n = orig[i]*4;
+ p[0] = palette[n ];
+ p[1] = palette[n+1];
+ p[2] = palette[n+2];
+ p[3] = palette[n+3];
+ p += 4;
+ }
+ }
+ STBI_FREE(a->out);
+ a->out = temp_out;
+
+ STBI_NOTUSED(len);
+
+ return 1;
+}
+
+static int stbi__unpremultiply_on_load = 0;
+static int stbi__de_iphone_flag = 0;
+
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply)
+{
+ stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply;
+}
+
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert)
+{
+ stbi__de_iphone_flag = flag_true_if_should_convert;
+}
+
+static void stbi__de_iphone(stbi__png *z)
+{
+ stbi__context *s = z->s;
+ stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+ stbi_uc *p = z->out;
+
+ if (s->img_out_n == 3) { // convert bgr to rgb
+ for (i=0; i < pixel_count; ++i) {
+ stbi_uc t = p[0];
+ p[0] = p[2];
+ p[2] = t;
+ p += 3;
+ }
+ } else {
+ STBI_ASSERT(s->img_out_n == 4);
+ if (stbi__unpremultiply_on_load) {
+ // convert bgr to rgb and unpremultiply
+ for (i=0; i < pixel_count; ++i) {
+ stbi_uc a = p[3];
+ stbi_uc t = p[0];
+ if (a) {
+ stbi_uc half = a / 2;
+ p[0] = (p[2] * 255 + half) / a;
+ p[1] = (p[1] * 255 + half) / a;
+ p[2] = ( t * 255 + half) / a;
+ } else {
+ p[0] = p[2];
+ p[2] = t;
+ }
+ p += 4;
+ }
+ } else {
+ // convert bgr to rgb
+ for (i=0; i < pixel_count; ++i) {
+ stbi_uc t = p[0];
+ p[0] = p[2];
+ p[2] = t;
+ p += 4;
+ }
+ }
+ }
+}
+
+#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d))
+
+static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)
+{
+ stbi_uc palette[1024], pal_img_n=0;
+ stbi_uc has_trans=0, tc[3]={0};
+ stbi__uint16 tc16[3];
+ stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0;
+ int first=1,k,interlace=0, color=0, is_iphone=0;
+ stbi__context *s = z->s;
+
+ z->expanded = NULL;
+ z->idata = NULL;
+ z->out = NULL;
+
+ if (!stbi__check_png_header(s)) return 0;
+
+ if (scan == STBI__SCAN_type) return 1;
+
+ for (;;) {
+ stbi__pngchunk c = stbi__get_chunk_header(s);
+ switch (c.type) {
+ case STBI__PNG_TYPE('C','g','B','I'):
+ is_iphone = 1;
+ stbi__skip(s, c.length);
+ break;
+ case STBI__PNG_TYPE('I','H','D','R'): {
+ int comp,filter;
+ if (!first) return stbi__err("multiple IHDR","Corrupt PNG");
+ first = 0;
+ if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG");
+ s->img_x = stbi__get32be(s); if (s->img_x > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");
+ s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");
+ z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only");
+ color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG");
+ if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG");
+ if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG");
+ comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG");
+ filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG");
+ interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG");
+ if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG");
+ if (!pal_img_n) {
+ s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);
+ if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode");
+ if (scan == STBI__SCAN_header) return 1;
+ } else {
+ // if paletted, then pal_n is our final components, and
+ // img_n is # components to decompress/filter.
+ s->img_n = 1;
+ if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG");
+ // if SCAN_header, have to scan to see if we have a tRNS
+ }
+ break;
+ }
+
+ case STBI__PNG_TYPE('P','L','T','E'): {
+ if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+ if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG");
+ pal_len = c.length / 3;
+ if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG");
+ for (i=0; i < pal_len; ++i) {
+ palette[i*4+0] = stbi__get8(s);
+ palette[i*4+1] = stbi__get8(s);
+ palette[i*4+2] = stbi__get8(s);
+ palette[i*4+3] = 255;
+ }
+ break;
+ }
+
+ case STBI__PNG_TYPE('t','R','N','S'): {
+ if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+ if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG");
+ if (pal_img_n) {
+ if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }
+ if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG");
+ if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG");
+ pal_img_n = 4;
+ for (i=0; i < c.length; ++i)
+ palette[i*4+3] = stbi__get8(s);
+ } else {
+ if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG");
+ if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG");
+ has_trans = 1;
+ if (z->depth == 16) {
+ for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is
+ } else {
+ for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger
+ }
+ }
+ break;
+ }
+
+ case STBI__PNG_TYPE('I','D','A','T'): {
+ if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+ if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG");
+ if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; }
+ if ((int)(ioff + c.length) < (int)ioff) return 0;
+ if (ioff + c.length > idata_limit) {
+ stbi__uint32 idata_limit_old = idata_limit;
+ stbi_uc *p;
+ if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;
+ while (ioff + c.length > idata_limit)
+ idata_limit *= 2;
+ STBI_NOTUSED(idata_limit_old);
+ p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory");
+ z->idata = p;
+ }
+ if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG");
+ ioff += c.length;
+ break;
+ }
+
+ case STBI__PNG_TYPE('I','E','N','D'): {
+ stbi__uint32 raw_len, bpl;
+ if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+ if (scan != STBI__SCAN_load) return 1;
+ if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG");
+ // initial guess for decoded data size to avoid unnecessary reallocs
+ bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component
+ raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;
+ z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);
+ if (z->expanded == NULL) return 0; // zlib should set error
+ STBI_FREE(z->idata); z->idata = NULL;
+ if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans)
+ s->img_out_n = s->img_n+1;
+ else
+ s->img_out_n = s->img_n;
+ if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0;
+ if (has_trans) {
+ if (z->depth == 16) {
+ if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0;
+ } else {
+ if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;
+ }
+ }
+ if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)
+ stbi__de_iphone(z);
+ if (pal_img_n) {
+ // pal_img_n == 3 or 4
+ s->img_n = pal_img_n; // record the actual colors we had
+ s->img_out_n = pal_img_n;
+ if (req_comp >= 3) s->img_out_n = req_comp;
+ if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))
+ return 0;
+ } else if (has_trans) {
+ // non-paletted image with tRNS -> source image has (constant) alpha
+ ++s->img_n;
+ }
+ STBI_FREE(z->expanded); z->expanded = NULL;
+ return 1;
+ }
+
+ default:
+ // if critical, fail
+ if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+ if ((c.type & (1 << 29)) == 0) {
+ #ifndef STBI_NO_FAILURE_STRINGS
+ // not threadsafe
+ static char invalid_chunk[] = "XXXX PNG chunk not known";
+ invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);
+ invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);
+ invalid_chunk[2] = STBI__BYTECAST(c.type >> 8);
+ invalid_chunk[3] = STBI__BYTECAST(c.type >> 0);
+ #endif
+ return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type");
+ }
+ stbi__skip(s, c.length);
+ break;
+ }
+ // end of PNG chunk, read and skip CRC
+ stbi__get32be(s);
+ }
+}
+
+static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri)
+{
+ void *result=NULL;
+ if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+ if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) {
+ if (p->depth < 8)
+ ri->bits_per_channel = 8;
+ else
+ ri->bits_per_channel = p->depth;
+ result = p->out;
+ p->out = NULL;
+ if (req_comp && req_comp != p->s->img_out_n) {
+ if (ri->bits_per_channel == 8)
+ result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+ else
+ result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+ p->s->img_out_n = req_comp;
+ if (result == NULL) return result;
+ }
+ *x = p->s->img_x;
+ *y = p->s->img_y;
+ if (n) *n = p->s->img_n;
+ }
+ STBI_FREE(p->out); p->out = NULL;
+ STBI_FREE(p->expanded); p->expanded = NULL;
+ STBI_FREE(p->idata); p->idata = NULL;
+
+ return result;
+}
+
+static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ stbi__png p;
+ p.s = s;
+ return stbi__do_png(&p, x,y,comp,req_comp, ri);
+}
+
+static int stbi__png_test(stbi__context *s)
+{
+ int r;
+ r = stbi__check_png_header(s);
+ stbi__rewind(s);
+ return r;
+}
+
+static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp)
+{
+ if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) {
+ stbi__rewind( p->s );
+ return 0;
+ }
+ if (x) *x = p->s->img_x;
+ if (y) *y = p->s->img_y;
+ if (comp) *comp = p->s->img_n;
+ return 1;
+}
+
+static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ stbi__png p;
+ p.s = s;
+ return stbi__png_info_raw(&p, x, y, comp);
+}
+
+static int stbi__png_is16(stbi__context *s)
+{
+ stbi__png p;
+ p.s = s;
+ if (!stbi__png_info_raw(&p, NULL, NULL, NULL))
+ return 0;
+ if (p.depth != 16) {
+ stbi__rewind(p.s);
+ return 0;
+ }
+ return 1;
+}
+#endif
+
+// Microsoft/Windows BMP image
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_test_raw(stbi__context *s)
+{
+ int r;
+ int sz;
+ if (stbi__get8(s) != 'B') return 0;
+ if (stbi__get8(s) != 'M') return 0;
+ stbi__get32le(s); // discard filesize
+ stbi__get16le(s); // discard reserved
+ stbi__get16le(s); // discard reserved
+ stbi__get32le(s); // discard data offset
+ sz = stbi__get32le(s);
+ r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124);
+ return r;
+}
+
+static int stbi__bmp_test(stbi__context *s)
+{
+ int r = stbi__bmp_test_raw(s);
+ stbi__rewind(s);
+ return r;
+}
+
+
+// returns 0..31 for the highest set bit
+static int stbi__high_bit(unsigned int z)
+{
+ int n=0;
+ if (z == 0) return -1;
+ if (z >= 0x10000) { n += 16; z >>= 16; }
+ if (z >= 0x00100) { n += 8; z >>= 8; }
+ if (z >= 0x00010) { n += 4; z >>= 4; }
+ if (z >= 0x00004) { n += 2; z >>= 2; }
+ if (z >= 0x00002) { n += 1;/* >>= 1;*/ }
+ return n;
+}
+
+static int stbi__bitcount(unsigned int a)
+{
+ a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2
+ a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4
+ a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits
+ a = (a + (a >> 8)); // max 16 per 8 bits
+ a = (a + (a >> 16)); // max 32 per 8 bits
+ return a & 0xff;
+}
+
+// extract an arbitrarily-aligned N-bit value (N=bits)
+// from v, and then make it 8-bits long and fractionally
+// extend it to full full range.
+static int stbi__shiftsigned(unsigned int v, int shift, int bits)
+{
+ static unsigned int mul_table[9] = {
+ 0,
+ 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/,
+ 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/,
+ };
+ static unsigned int shift_table[9] = {
+ 0, 0,0,1,0,2,4,6,0,
+ };
+ if (shift < 0)
+ v <<= -shift;
+ else
+ v >>= shift;
+ STBI_ASSERT(v >= 0 && v < 256);
+ v >>= (8-bits);
+ STBI_ASSERT(bits >= 0 && bits <= 8);
+ return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits];
+}
+
+typedef struct
+{
+ int bpp, offset, hsz;
+ unsigned int mr,mg,mb,ma, all_a;
+} stbi__bmp_data;
+
+static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info)
+{
+ int hsz;
+ if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP");
+ stbi__get32le(s); // discard filesize
+ stbi__get16le(s); // discard reserved
+ stbi__get16le(s); // discard reserved
+ info->offset = stbi__get32le(s);
+ info->hsz = hsz = stbi__get32le(s);
+ info->mr = info->mg = info->mb = info->ma = 0;
+
+ if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown");
+ if (hsz == 12) {
+ s->img_x = stbi__get16le(s);
+ s->img_y = stbi__get16le(s);
+ } else {
+ s->img_x = stbi__get32le(s);
+ s->img_y = stbi__get32le(s);
+ }
+ if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP");
+ info->bpp = stbi__get16le(s);
+ if (hsz != 12) {
+ int compress = stbi__get32le(s);
+ if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE");
+ stbi__get32le(s); // discard sizeof
+ stbi__get32le(s); // discard hres
+ stbi__get32le(s); // discard vres
+ stbi__get32le(s); // discard colorsused
+ stbi__get32le(s); // discard max important
+ if (hsz == 40 || hsz == 56) {
+ if (hsz == 56) {
+ stbi__get32le(s);
+ stbi__get32le(s);
+ stbi__get32le(s);
+ stbi__get32le(s);
+ }
+ if (info->bpp == 16 || info->bpp == 32) {
+ if (compress == 0) {
+ if (info->bpp == 32) {
+ info->mr = 0xffu << 16;
+ info->mg = 0xffu << 8;
+ info->mb = 0xffu << 0;
+ info->ma = 0xffu << 24;
+ info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0
+ } else {
+ info->mr = 31u << 10;
+ info->mg = 31u << 5;
+ info->mb = 31u << 0;
+ }
+ } else if (compress == 3) {
+ info->mr = stbi__get32le(s);
+ info->mg = stbi__get32le(s);
+ info->mb = stbi__get32le(s);
+ // not documented, but generated by photoshop and handled by mspaint
+ if (info->mr == info->mg && info->mg == info->mb) {
+ // ?!?!?
+ return stbi__errpuc("bad BMP", "bad BMP");
+ }
+ } else
+ return stbi__errpuc("bad BMP", "bad BMP");
+ }
+ } else {
+ int i;
+ if (hsz != 108 && hsz != 124)
+ return stbi__errpuc("bad BMP", "bad BMP");
+ info->mr = stbi__get32le(s);
+ info->mg = stbi__get32le(s);
+ info->mb = stbi__get32le(s);
+ info->ma = stbi__get32le(s);
+ stbi__get32le(s); // discard color space
+ for (i=0; i < 12; ++i)
+ stbi__get32le(s); // discard color space parameters
+ if (hsz == 124) {
+ stbi__get32le(s); // discard rendering intent
+ stbi__get32le(s); // discard offset of profile data
+ stbi__get32le(s); // discard size of profile data
+ stbi__get32le(s); // discard reserved
+ }
+ }
+ }
+ return (void *) 1;
+}
+
+
+static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ stbi_uc *out;
+ unsigned int mr=0,mg=0,mb=0,ma=0, all_a;
+ stbi_uc pal[256][4];
+ int psize=0,i,j,width;
+ int flip_vertically, pad, target;
+ stbi__bmp_data info;
+ STBI_NOTUSED(ri);
+
+ info.all_a = 255;
+ if (stbi__bmp_parse_header(s, &info) == NULL)
+ return NULL; // error code already set
+
+ flip_vertically = ((int) s->img_y) > 0;
+ s->img_y = abs((int) s->img_y);
+
+ mr = info.mr;
+ mg = info.mg;
+ mb = info.mb;
+ ma = info.ma;
+ all_a = info.all_a;
+
+ if (info.hsz == 12) {
+ if (info.bpp < 24)
+ psize = (info.offset - 14 - 24) / 3;
+ } else {
+ if (info.bpp < 16)
+ psize = (info.offset - 14 - info.hsz) >> 2;
+ }
+
+ if (info.bpp == 24 && ma == 0xff000000)
+ s->img_n = 3;
+ else
+ s->img_n = ma ? 4 : 3;
+ if (req_comp && req_comp >= 3) // we can directly decode 3 or 4
+ target = req_comp;
+ else
+ target = s->img_n; // if they want monochrome, we'll post-convert
+
+ // sanity-check size
+ if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0))
+ return stbi__errpuc("too large", "Corrupt BMP");
+
+ out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0);
+ if (!out) return stbi__errpuc("outofmem", "Out of memory");
+ if (info.bpp < 16) {
+ int z=0;
+ if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); }
+ for (i=0; i < psize; ++i) {
+ pal[i][2] = stbi__get8(s);
+ pal[i][1] = stbi__get8(s);
+ pal[i][0] = stbi__get8(s);
+ if (info.hsz != 12) stbi__get8(s);
+ pal[i][3] = 255;
+ }
+ stbi__skip(s, info.offset - 14 - info.hsz - psize * (info.hsz == 12 ? 3 : 4));
+ if (info.bpp == 1) width = (s->img_x + 7) >> 3;
+ else if (info.bpp == 4) width = (s->img_x + 1) >> 1;
+ else if (info.bpp == 8) width = s->img_x;
+ else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); }
+ pad = (-width)&3;
+ if (info.bpp == 1) {
+ for (j=0; j < (int) s->img_y; ++j) {
+ int bit_offset = 7, v = stbi__get8(s);
+ for (i=0; i < (int) s->img_x; ++i) {
+ int color = (v>>bit_offset)&0x1;
+ out[z++] = pal[color][0];
+ out[z++] = pal[color][1];
+ out[z++] = pal[color][2];
+ if (target == 4) out[z++] = 255;
+ if (i+1 == (int) s->img_x) break;
+ if((--bit_offset) < 0) {
+ bit_offset = 7;
+ v = stbi__get8(s);
+ }
+ }
+ stbi__skip(s, pad);
+ }
+ } else {
+ for (j=0; j < (int) s->img_y; ++j) {
+ for (i=0; i < (int) s->img_x; i += 2) {
+ int v=stbi__get8(s),v2=0;
+ if (info.bpp == 4) {
+ v2 = v & 15;
+ v >>= 4;
+ }
+ out[z++] = pal[v][0];
+ out[z++] = pal[v][1];
+ out[z++] = pal[v][2];
+ if (target == 4) out[z++] = 255;
+ if (i+1 == (int) s->img_x) break;
+ v = (info.bpp == 8) ? stbi__get8(s) : v2;
+ out[z++] = pal[v][0];
+ out[z++] = pal[v][1];
+ out[z++] = pal[v][2];
+ if (target == 4) out[z++] = 255;
+ }
+ stbi__skip(s, pad);
+ }
+ }
+ } else {
+ int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0;
+ int z = 0;
+ int easy=0;
+ stbi__skip(s, info.offset - 14 - info.hsz);
+ if (info.bpp == 24) width = 3 * s->img_x;
+ else if (info.bpp == 16) width = 2*s->img_x;
+ else /* bpp = 32 and pad = 0 */ width=0;
+ pad = (-width) & 3;
+ if (info.bpp == 24) {
+ easy = 1;
+ } else if (info.bpp == 32) {
+ if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000)
+ easy = 2;
+ }
+ if (!easy) {
+ if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); }
+ // right shift amt to put high bit in position #7
+ rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr);
+ gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg);
+ bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb);
+ ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma);
+ }
+ for (j=0; j < (int) s->img_y; ++j) {
+ if (easy) {
+ for (i=0; i < (int) s->img_x; ++i) {
+ unsigned char a;
+ out[z+2] = stbi__get8(s);
+ out[z+1] = stbi__get8(s);
+ out[z+0] = stbi__get8(s);
+ z += 3;
+ a = (easy == 2 ? stbi__get8(s) : 255);
+ all_a |= a;
+ if (target == 4) out[z++] = a;
+ }
+ } else {
+ int bpp = info.bpp;
+ for (i=0; i < (int) s->img_x; ++i) {
+ stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s));
+ unsigned int a;
+ out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount));
+ out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount));
+ out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount));
+ a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255);
+ all_a |= a;
+ if (target == 4) out[z++] = STBI__BYTECAST(a);
+ }
+ }
+ stbi__skip(s, pad);
+ }
+ }
+
+ // if alpha channel is all 0s, replace with all 255s
+ if (target == 4 && all_a == 0)
+ for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4)
+ out[i] = 255;
+
+ if (flip_vertically) {
+ stbi_uc t;
+ for (j=0; j < (int) s->img_y>>1; ++j) {
+ stbi_uc *p1 = out + j *s->img_x*target;
+ stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target;
+ for (i=0; i < (int) s->img_x*target; ++i) {
+ t = p1[i]; p1[i] = p2[i]; p2[i] = t;
+ }
+ }
+ }
+
+ if (req_comp && req_comp != target) {
+ out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y);
+ if (out == NULL) return out; // stbi__convert_format frees input on failure
+ }
+
+ *x = s->img_x;
+ *y = s->img_y;
+ if (comp) *comp = s->img_n;
+ return out;
+}
+#endif
+
+// Targa Truevision - TGA
+// by Jonathan Dummer
+#ifndef STBI_NO_TGA
+// returns STBI_rgb or whatever, 0 on error
+static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16)
+{
+ // only RGB or RGBA (incl. 16bit) or grey allowed
+ if (is_rgb16) *is_rgb16 = 0;
+ switch(bits_per_pixel) {
+ case 8: return STBI_grey;
+ case 16: if(is_grey) return STBI_grey_alpha;
+ // fallthrough
+ case 15: if(is_rgb16) *is_rgb16 = 1;
+ return STBI_rgb;
+ case 24: // fallthrough
+ case 32: return bits_per_pixel/8;
+ default: return 0;
+ }
+}
+
+static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp;
+ int sz, tga_colormap_type;
+ stbi__get8(s); // discard Offset
+ tga_colormap_type = stbi__get8(s); // colormap type
+ if( tga_colormap_type > 1 ) {
+ stbi__rewind(s);
+ return 0; // only RGB or indexed allowed
+ }
+ tga_image_type = stbi__get8(s); // image type
+ if ( tga_colormap_type == 1 ) { // colormapped (paletted) image
+ if (tga_image_type != 1 && tga_image_type != 9) {
+ stbi__rewind(s);
+ return 0;
+ }
+ stbi__skip(s,4); // skip index of first colormap entry and number of entries
+ sz = stbi__get8(s); // check bits per palette color entry
+ if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) {
+ stbi__rewind(s);
+ return 0;
+ }
+ stbi__skip(s,4); // skip image x and y origin
+ tga_colormap_bpp = sz;
+ } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE
+ if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) {
+ stbi__rewind(s);
+ return 0; // only RGB or grey allowed, +/- RLE
+ }
+ stbi__skip(s,9); // skip colormap specification and image x/y origin
+ tga_colormap_bpp = 0;
+ }
+ tga_w = stbi__get16le(s);
+ if( tga_w < 1 ) {
+ stbi__rewind(s);
+ return 0; // test width
+ }
+ tga_h = stbi__get16le(s);
+ if( tga_h < 1 ) {
+ stbi__rewind(s);
+ return 0; // test height
+ }
+ tga_bits_per_pixel = stbi__get8(s); // bits per pixel
+ stbi__get8(s); // ignore alpha bits
+ if (tga_colormap_bpp != 0) {
+ if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) {
+ // when using a colormap, tga_bits_per_pixel is the size of the indexes
+ // I don't think anything but 8 or 16bit indexes makes sense
+ stbi__rewind(s);
+ return 0;
+ }
+ tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL);
+ } else {
+ tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL);
+ }
+ if(!tga_comp) {
+ stbi__rewind(s);
+ return 0;
+ }
+ if (x) *x = tga_w;
+ if (y) *y = tga_h;
+ if (comp) *comp = tga_comp;
+ return 1; // seems to have passed everything
+}
+
+static int stbi__tga_test(stbi__context *s)
+{
+ int res = 0;
+ int sz, tga_color_type;
+ stbi__get8(s); // discard Offset
+ tga_color_type = stbi__get8(s); // color type
+ if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed
+ sz = stbi__get8(s); // image type
+ if ( tga_color_type == 1 ) { // colormapped (paletted) image
+ if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9
+ stbi__skip(s,4); // skip index of first colormap entry and number of entries
+ sz = stbi__get8(s); // check bits per palette color entry
+ if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+ stbi__skip(s,4); // skip image x and y origin
+ } else { // "normal" image w/o colormap
+ if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE
+ stbi__skip(s,9); // skip colormap specification and image x/y origin
+ }
+ if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width
+ if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height
+ sz = stbi__get8(s); // bits per pixel
+ if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index
+ if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+
+ res = 1; // if we got this far, everything's good and we can return 1 instead of 0
+
+errorEnd:
+ stbi__rewind(s);
+ return res;
+}
+
+// read 16bit value and convert to 24bit RGB
+static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out)
+{
+ stbi__uint16 px = (stbi__uint16)stbi__get16le(s);
+ stbi__uint16 fiveBitMask = 31;
+ // we have 3 channels with 5bits each
+ int r = (px >> 10) & fiveBitMask;
+ int g = (px >> 5) & fiveBitMask;
+ int b = px & fiveBitMask;
+ // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later
+ out[0] = (stbi_uc)((r * 255)/31);
+ out[1] = (stbi_uc)((g * 255)/31);
+ out[2] = (stbi_uc)((b * 255)/31);
+
+ // some people claim that the most significant bit might be used for alpha
+ // (possibly if an alpha-bit is set in the "image descriptor byte")
+ // but that only made 16bit test images completely translucent..
+ // so let's treat all 15 and 16bit TGAs as RGB with no alpha.
+}
+
+static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ // read in the TGA header stuff
+ int tga_offset = stbi__get8(s);
+ int tga_indexed = stbi__get8(s);
+ int tga_image_type = stbi__get8(s);
+ int tga_is_RLE = 0;
+ int tga_palette_start = stbi__get16le(s);
+ int tga_palette_len = stbi__get16le(s);
+ int tga_palette_bits = stbi__get8(s);
+ int tga_x_origin = stbi__get16le(s);
+ int tga_y_origin = stbi__get16le(s);
+ int tga_width = stbi__get16le(s);
+ int tga_height = stbi__get16le(s);
+ int tga_bits_per_pixel = stbi__get8(s);
+ int tga_comp, tga_rgb16=0;
+ int tga_inverted = stbi__get8(s);
+ // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?)
+ // image data
+ unsigned char *tga_data;
+ unsigned char *tga_palette = NULL;
+ int i, j;
+ unsigned char raw_data[4] = {0};
+ int RLE_count = 0;
+ int RLE_repeating = 0;
+ int read_next_pixel = 1;
+ STBI_NOTUSED(ri);
+ STBI_NOTUSED(tga_x_origin); // @TODO
+ STBI_NOTUSED(tga_y_origin); // @TODO
+
+ // do a tiny bit of precessing
+ if ( tga_image_type >= 8 )
+ {
+ tga_image_type -= 8;
+ tga_is_RLE = 1;
+ }
+ tga_inverted = 1 - ((tga_inverted >> 5) & 1);
+
+ // If I'm paletted, then I'll use the number of bits from the palette
+ if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16);
+ else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16);
+
+ if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency
+ return stbi__errpuc("bad format", "Can't find out TGA pixelformat");
+
+ // tga info
+ *x = tga_width;
+ *y = tga_height;
+ if (comp) *comp = tga_comp;
+
+ if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0))
+ return stbi__errpuc("too large", "Corrupt TGA");
+
+ tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0);
+ if (!tga_data) return stbi__errpuc("outofmem", "Out of memory");
+
+ // skip to the data's starting position (offset usually = 0)
+ stbi__skip(s, tga_offset );
+
+ if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) {
+ for (i=0; i < tga_height; ++i) {
+ int row = tga_inverted ? tga_height -i - 1 : i;
+ stbi_uc *tga_row = tga_data + row*tga_width*tga_comp;
+ stbi__getn(s, tga_row, tga_width * tga_comp);
+ }
+ } else {
+ // do I need to load a palette?
+ if ( tga_indexed)
+ {
+ // any data to skip? (offset usually = 0)
+ stbi__skip(s, tga_palette_start );
+ // load the palette
+ tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0);
+ if (!tga_palette) {
+ STBI_FREE(tga_data);
+ return stbi__errpuc("outofmem", "Out of memory");
+ }
+ if (tga_rgb16) {
+ stbi_uc *pal_entry = tga_palette;
+ STBI_ASSERT(tga_comp == STBI_rgb);
+ for (i=0; i < tga_palette_len; ++i) {
+ stbi__tga_read_rgb16(s, pal_entry);
+ pal_entry += tga_comp;
+ }
+ } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) {
+ STBI_FREE(tga_data);
+ STBI_FREE(tga_palette);
+ return stbi__errpuc("bad palette", "Corrupt TGA");
+ }
+ }
+ // load the data
+ for (i=0; i < tga_width * tga_height; ++i)
+ {
+ // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk?
+ if ( tga_is_RLE )
+ {
+ if ( RLE_count == 0 )
+ {
+ // yep, get the next byte as a RLE command
+ int RLE_cmd = stbi__get8(s);
+ RLE_count = 1 + (RLE_cmd & 127);
+ RLE_repeating = RLE_cmd >> 7;
+ read_next_pixel = 1;
+ } else if ( !RLE_repeating )
+ {
+ read_next_pixel = 1;
+ }
+ } else
+ {
+ read_next_pixel = 1;
+ }
+ // OK, if I need to read a pixel, do it now
+ if ( read_next_pixel )
+ {
+ // load however much data we did have
+ if ( tga_indexed )
+ {
+ // read in index, then perform the lookup
+ int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s);
+ if ( pal_idx >= tga_palette_len ) {
+ // invalid index
+ pal_idx = 0;
+ }
+ pal_idx *= tga_comp;
+ for (j = 0; j < tga_comp; ++j) {
+ raw_data[j] = tga_palette[pal_idx+j];
+ }
+ } else if(tga_rgb16) {
+ STBI_ASSERT(tga_comp == STBI_rgb);
+ stbi__tga_read_rgb16(s, raw_data);
+ } else {
+ // read in the data raw
+ for (j = 0; j < tga_comp; ++j) {
+ raw_data[j] = stbi__get8(s);
+ }
+ }
+ // clear the reading flag for the next pixel
+ read_next_pixel = 0;
+ } // end of reading a pixel
+
+ // copy data
+ for (j = 0; j < tga_comp; ++j)
+ tga_data[i*tga_comp+j] = raw_data[j];
+
+ // in case we're in RLE mode, keep counting down
+ --RLE_count;
+ }
+ // do I need to invert the image?
+ if ( tga_inverted )
+ {
+ for (j = 0; j*2 < tga_height; ++j)
+ {
+ int index1 = j * tga_width * tga_comp;
+ int index2 = (tga_height - 1 - j) * tga_width * tga_comp;
+ for (i = tga_width * tga_comp; i > 0; --i)
+ {
+ unsigned char temp = tga_data[index1];
+ tga_data[index1] = tga_data[index2];
+ tga_data[index2] = temp;
+ ++index1;
+ ++index2;
+ }
+ }
+ }
+ // clear my palette, if I had one
+ if ( tga_palette != NULL )
+ {
+ STBI_FREE( tga_palette );
+ }
+ }
+
+ // swap RGB - if the source data was RGB16, it already is in the right order
+ if (tga_comp >= 3 && !tga_rgb16)
+ {
+ unsigned char* tga_pixel = tga_data;
+ for (i=0; i < tga_width * tga_height; ++i)
+ {
+ unsigned char temp = tga_pixel[0];
+ tga_pixel[0] = tga_pixel[2];
+ tga_pixel[2] = temp;
+ tga_pixel += tga_comp;
+ }
+ }
+
+ // convert to target component count
+ if (req_comp && req_comp != tga_comp)
+ tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height);
+
+ // the things I do to get rid of an error message, and yet keep
+ // Microsoft's C compilers happy... [8^(
+ tga_palette_start = tga_palette_len = tga_palette_bits =
+ tga_x_origin = tga_y_origin = 0;
+ STBI_NOTUSED(tga_palette_start);
+ // OK, done
+ return tga_data;
+}
+#endif
+
+// *************************************************************************************************
+// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_test(stbi__context *s)
+{
+ int r = (stbi__get32be(s) == 0x38425053);
+ stbi__rewind(s);
+ return r;
+}
+
+static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount)
+{
+ int count, nleft, len;
+
+ count = 0;
+ while ((nleft = pixelCount - count) > 0) {
+ len = stbi__get8(s);
+ if (len == 128) {
+ // No-op.
+ } else if (len < 128) {
+ // Copy next len+1 bytes literally.
+ len++;
+ if (len > nleft) return 0; // corrupt data
+ count += len;
+ while (len) {
+ *p = stbi__get8(s);
+ p += 4;
+ len--;
+ }
+ } else if (len > 128) {
+ stbi_uc val;
+ // Next -len+1 bytes in the dest are replicated from next source byte.
+ // (Interpret len as a negative 8-bit int.)
+ len = 257 - len;
+ if (len > nleft) return 0; // corrupt data
+ val = stbi__get8(s);
+ count += len;
+ while (len) {
+ *p = val;
+ p += 4;
+ len--;
+ }
+ }
+ }
+
+ return 1;
+}
+
+static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+ int pixelCount;
+ int channelCount, compression;
+ int channel, i;
+ int bitdepth;
+ int w,h;
+ stbi_uc *out;
+ STBI_NOTUSED(ri);
+
+ // Check identifier
+ if (stbi__get32be(s) != 0x38425053) // "8BPS"
+ return stbi__errpuc("not PSD", "Corrupt PSD image");
+
+ // Check file type version.
+ if (stbi__get16be(s) != 1)
+ return stbi__errpuc("wrong version", "Unsupported version of PSD image");
+
+ // Skip 6 reserved bytes.
+ stbi__skip(s, 6 );
+
+ // Read the number of channels (R, G, B, A, etc).
+ channelCount = stbi__get16be(s);
+ if (channelCount < 0 || channelCount > 16)
+ return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image");
+
+ // Read the rows and columns of the image.
+ h = stbi__get32be(s);
+ w = stbi__get32be(s);
+
+ // Make sure the depth is 8 bits.
+ bitdepth = stbi__get16be(s);
+ if (bitdepth != 8 && bitdepth != 16)
+ return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit");
+
+ // Make sure the color mode is RGB.
+ // Valid options are:
+ // 0: Bitmap
+ // 1: Grayscale
+ // 2: Indexed color
+ // 3: RGB color
+ // 4: CMYK color
+ // 7: Multichannel
+ // 8: Duotone
+ // 9: Lab color
+ if (stbi__get16be(s) != 3)
+ return stbi__errpuc("wrong color format", "PSD is not in RGB color format");
+
+ // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.)
+ stbi__skip(s,stbi__get32be(s) );
+
+ // Skip the image resources. (resolution, pen tool paths, etc)
+ stbi__skip(s, stbi__get32be(s) );
+
+ // Skip the reserved data.
+ stbi__skip(s, stbi__get32be(s) );
+
+ // Find out if the data is compressed.
+ // Known values:
+ // 0: no compression
+ // 1: RLE compressed
+ compression = stbi__get16be(s);
+ if (compression > 1)
+ return stbi__errpuc("bad compression", "PSD has an unknown compression format");
+
+ // Check size
+ if (!stbi__mad3sizes_valid(4, w, h, 0))
+ return stbi__errpuc("too large", "Corrupt PSD");
+
+ // Create the destination image.
+
+ if (!compression && bitdepth == 16 && bpc == 16) {
+ out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0);
+ ri->bits_per_channel = 16;
+ } else
+ out = (stbi_uc *) stbi__malloc(4 * w*h);
+
+ if (!out) return stbi__errpuc("outofmem", "Out of memory");
+ pixelCount = w*h;
+
+ // Initialize the data to zero.
+ //memset( out, 0, pixelCount * 4 );
+
+ // Finally, the image data.
+ if (compression) {
+ // RLE as used by .PSD and .TIFF
+ // Loop until you get the number of unpacked bytes you are expecting:
+ // Read the next source byte into n.
+ // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally.
+ // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times.
+ // Else if n is 128, noop.
+ // Endloop
+
+ // The RLE-compressed data is preceded by a 2-byte data count for each row in the data,
+ // which we're going to just skip.
+ stbi__skip(s, h * channelCount * 2 );
+
+ // Read the RLE data by channel.
+ for (channel = 0; channel < 4; channel++) {
+ stbi_uc *p;
+
+ p = out+channel;
+ if (channel >= channelCount) {
+ // Fill this channel with default data.
+ for (i = 0; i < pixelCount; i++, p += 4)
+ *p = (channel == 3 ? 255 : 0);
+ } else {
+ // Read the RLE data.
+ if (!stbi__psd_decode_rle(s, p, pixelCount)) {
+ STBI_FREE(out);
+ return stbi__errpuc("corrupt", "bad RLE data");
+ }
+ }
+ }
+
+ } else {
+ // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...)
+ // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image.
+
+ // Read the data by channel.
+ for (channel = 0; channel < 4; channel++) {
+ if (channel >= channelCount) {
+ // Fill this channel with default data.
+ if (bitdepth == 16 && bpc == 16) {
+ stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+ stbi__uint16 val = channel == 3 ? 65535 : 0;
+ for (i = 0; i < pixelCount; i++, q += 4)
+ *q = val;
+ } else {
+ stbi_uc *p = out+channel;
+ stbi_uc val = channel == 3 ? 255 : 0;
+ for (i = 0; i < pixelCount; i++, p += 4)
+ *p = val;
+ }
+ } else {
+ if (ri->bits_per_channel == 16) { // output bpc
+ stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+ for (i = 0; i < pixelCount; i++, q += 4)
+ *q = (stbi__uint16) stbi__get16be(s);
+ } else {
+ stbi_uc *p = out+channel;
+ if (bitdepth == 16) { // input bpc
+ for (i = 0; i < pixelCount; i++, p += 4)
+ *p = (stbi_uc) (stbi__get16be(s) >> 8);
+ } else {
+ for (i = 0; i < pixelCount; i++, p += 4)
+ *p = stbi__get8(s);
+ }
+ }
+ }
+ }
+ }
+
+ // remove weird white matte from PSD
+ if (channelCount >= 4) {
+ if (ri->bits_per_channel == 16) {
+ for (i=0; i < w*h; ++i) {
+ stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i;
+ if (pixel[3] != 0 && pixel[3] != 65535) {
+ float a = pixel[3] / 65535.0f;
+ float ra = 1.0f / a;
+ float inv_a = 65535.0f * (1 - ra);
+ pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a);
+ pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a);
+ pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a);
+ }
+ }
+ } else {
+ for (i=0; i < w*h; ++i) {
+ unsigned char *pixel = out + 4*i;
+ if (pixel[3] != 0 && pixel[3] != 255) {
+ float a = pixel[3] / 255.0f;
+ float ra = 1.0f / a;
+ float inv_a = 255.0f * (1 - ra);
+ pixel[0] = (unsigned char) (pixel[0]*ra + inv_a);
+ pixel[1] = (unsigned char) (pixel[1]*ra + inv_a);
+ pixel[2] = (unsigned char) (pixel[2]*ra + inv_a);
+ }
+ }
+ }
+ }
+
+ // convert to desired output format
+ if (req_comp && req_comp != 4) {
+ if (ri->bits_per_channel == 16)
+ out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h);
+ else
+ out = stbi__convert_format(out, 4, req_comp, w, h);
+ if (out == NULL) return out; // stbi__convert_format frees input on failure
+ }
+
+ if (comp) *comp = 4;
+ *y = h;
+ *x = w;
+
+ return out;
+}
+#endif
+
+// *************************************************************************************************
+// Softimage PIC loader
+// by Tom Seddon
+//
+// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format
+// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_is4(stbi__context *s,const char *str)
+{
+ int i;
+ for (i=0; i<4; ++i)
+ if (stbi__get8(s) != (stbi_uc)str[i])
+ return 0;
+
+ return 1;
+}
+
+static int stbi__pic_test_core(stbi__context *s)
+{
+ int i;
+
+ if (!stbi__pic_is4(s,"\x53\x80\xF6\x34"))
+ return 0;
+
+ for(i=0;i<84;++i)
+ stbi__get8(s);
+
+ if (!stbi__pic_is4(s,"PICT"))
+ return 0;
+
+ return 1;
+}
+
+typedef struct
+{
+ stbi_uc size,type,channel;
+} stbi__pic_packet;
+
+static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest)
+{
+ int mask=0x80, i;
+
+ for (i=0; i<4; ++i, mask>>=1) {
+ if (channel & mask) {
+ if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short");
+ dest[i]=stbi__get8(s);
+ }
+ }
+
+ return dest;
+}
+
+static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src)
+{
+ int mask=0x80,i;
+
+ for (i=0;i<4; ++i, mask>>=1)
+ if (channel&mask)
+ dest[i]=src[i];
+}
+
+static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result)
+{
+ int act_comp=0,num_packets=0,y,chained;
+ stbi__pic_packet packets[10];
+
+ // this will (should...) cater for even some bizarre stuff like having data
+ // for the same channel in multiple packets.
+ do {
+ stbi__pic_packet *packet;
+
+ if (num_packets==sizeof(packets)/sizeof(packets[0]))
+ return stbi__errpuc("bad format","too many packets");
+
+ packet = &packets[num_packets++];
+
+ chained = stbi__get8(s);
+ packet->size = stbi__get8(s);
+ packet->type = stbi__get8(s);
+ packet->channel = stbi__get8(s);
+
+ act_comp |= packet->channel;
+
+ if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)");
+ if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp");
+ } while (chained);
+
+ *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel?
+
+ for(y=0; y<height; ++y) {
+ int packet_idx;
+
+ for(packet_idx=0; packet_idx < num_packets; ++packet_idx) {
+ stbi__pic_packet *packet = &packets[packet_idx];
+ stbi_uc *dest = result+y*width*4;
+
+ switch (packet->type) {
+ default:
+ return stbi__errpuc("bad format","packet has bad compression type");
+
+ case 0: {//uncompressed
+ int x;
+
+ for(x=0;x<width;++x, dest+=4)
+ if (!stbi__readval(s,packet->channel,dest))
+ return 0;
+ break;
+ }
+
+ case 1://Pure RLE
+ {
+ int left=width, i;
+
+ while (left>0) {
+ stbi_uc count,value[4];
+
+ count=stbi__get8(s);
+ if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)");
+
+ if (count > left)
+ count = (stbi_uc) left;
+
+ if (!stbi__readval(s,packet->channel,value)) return 0;
+
+ for(i=0; i<count; ++i,dest+=4)
+ stbi__copyval(packet->channel,dest,value);
+ left -= count;
+ }
+ }
+ break;
+
+ case 2: {//Mixed RLE
+ int left=width;
+ while (left>0) {
+ int count = stbi__get8(s), i;
+ if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)");
+
+ if (count >= 128) { // Repeated
+ stbi_uc value[4];
+
+ if (count==128)
+ count = stbi__get16be(s);
+ else
+ count -= 127;
+ if (count > left)
+ return stbi__errpuc("bad file","scanline overrun");
+
+ if (!stbi__readval(s,packet->channel,value))
+ return 0;
+
+ for(i=0;i<count;++i, dest += 4)
+ stbi__copyval(packet->channel,dest,value);
+ } else { // Raw
+ ++count;
+ if (count>left) return stbi__errpuc("bad file","scanline overrun");
+
+ for(i=0;i<count;++i, dest+=4)
+ if (!stbi__readval(s,packet->channel,dest))
+ return 0;
+ }
+ left-=count;
+ }
+ break;
+ }
+ }
+ }
+ }
+
+ return result;
+}
+
+static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri)
+{
+ stbi_uc *result;
+ int i, x,y, internal_comp;
+ STBI_NOTUSED(ri);
+
+ if (!comp) comp = &internal_comp;
+
+ for (i=0; i<92; ++i)
+ stbi__get8(s);
+
+ x = stbi__get16be(s);
+ y = stbi__get16be(s);
+ if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)");
+ if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode");
+
+ stbi__get32be(s); //skip `ratio'
+ stbi__get16be(s); //skip `fields'
+ stbi__get16be(s); //skip `pad'
+
+ // intermediate buffer is RGBA
+ result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0);
+ memset(result, 0xff, x*y*4);
+
+ if (!stbi__pic_load_core(s,x,y,comp, result)) {
+ STBI_FREE(result);
+ result=0;
+ }
+ *px = x;
+ *py = y;
+ if (req_comp == 0) req_comp = *comp;
+ result=stbi__convert_format(result,4,req_comp,x,y);
+
+ return result;
+}
+
+static int stbi__pic_test(stbi__context *s)
+{
+ int r = stbi__pic_test_core(s);
+ stbi__rewind(s);
+ return r;
+}
+#endif
+
+// *************************************************************************************************
+// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb
+
+#ifndef STBI_NO_GIF
+typedef struct
+{
+ stbi__int16 prefix;
+ stbi_uc first;
+ stbi_uc suffix;
+} stbi__gif_lzw;
+
+typedef struct
+{
+ int w,h;
+ stbi_uc *out; // output buffer (always 4 components)
+ stbi_uc *background; // The current "background" as far as a gif is concerned
+ stbi_uc *history;
+ int flags, bgindex, ratio, transparent, eflags;
+ stbi_uc pal[256][4];
+ stbi_uc lpal[256][4];
+ stbi__gif_lzw codes[8192];
+ stbi_uc *color_table;
+ int parse, step;
+ int lflags;
+ int start_x, start_y;
+ int max_x, max_y;
+ int cur_x, cur_y;
+ int line_size;
+ int delay;
+} stbi__gif;
+
+static int stbi__gif_test_raw(stbi__context *s)
+{
+ int sz;
+ if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0;
+ sz = stbi__get8(s);
+ if (sz != '9' && sz != '7') return 0;
+ if (stbi__get8(s) != 'a') return 0;
+ return 1;
+}
+
+static int stbi__gif_test(stbi__context *s)
+{
+ int r = stbi__gif_test_raw(s);
+ stbi__rewind(s);
+ return r;
+}
+
+static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp)
+{
+ int i;
+ for (i=0; i < num_entries; ++i) {
+ pal[i][2] = stbi__get8(s);
+ pal[i][1] = stbi__get8(s);
+ pal[i][0] = stbi__get8(s);
+ pal[i][3] = transp == i ? 0 : 255;
+ }
+}
+
+static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info)
+{
+ stbi_uc version;
+ if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8')
+ return stbi__err("not GIF", "Corrupt GIF");
+
+ version = stbi__get8(s);
+ if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF");
+ if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF");
+
+ stbi__g_failure_reason = "";
+ g->w = stbi__get16le(s);
+ g->h = stbi__get16le(s);
+ g->flags = stbi__get8(s);
+ g->bgindex = stbi__get8(s);
+ g->ratio = stbi__get8(s);
+ g->transparent = -1;
+
+ if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments
+
+ if (is_info) return 1;
+
+ if (g->flags & 0x80)
+ stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1);
+
+ return 1;
+}
+
+static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp)
+{
+ stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif));
+ if (!stbi__gif_header(s, g, comp, 1)) {
+ STBI_FREE(g);
+ stbi__rewind( s );
+ return 0;
+ }
+ if (x) *x = g->w;
+ if (y) *y = g->h;
+ STBI_FREE(g);
+ return 1;
+}
+
+static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code)
+{
+ stbi_uc *p, *c;
+ int idx;
+
+ // recurse to decode the prefixes, since the linked-list is backwards,
+ // and working backwards through an interleaved image would be nasty
+ if (g->codes[code].prefix >= 0)
+ stbi__out_gif_code(g, g->codes[code].prefix);
+
+ if (g->cur_y >= g->max_y) return;
+
+ idx = g->cur_x + g->cur_y;
+ p = &g->out[idx];
+ g->history[idx / 4] = 1;
+
+ c = &g->color_table[g->codes[code].suffix * 4];
+ if (c[3] > 128) { // don't render transparent pixels;
+ p[0] = c[2];
+ p[1] = c[1];
+ p[2] = c[0];
+ p[3] = c[3];
+ }
+ g->cur_x += 4;
+
+ if (g->cur_x >= g->max_x) {
+ g->cur_x = g->start_x;
+ g->cur_y += g->step;
+
+ while (g->cur_y >= g->max_y && g->parse > 0) {
+ g->step = (1 << g->parse) * g->line_size;
+ g->cur_y = g->start_y + (g->step >> 1);
+ --g->parse;
+ }
+ }
+}
+
+static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g)
+{
+ stbi_uc lzw_cs;
+ stbi__int32 len, init_code;
+ stbi__uint32 first;
+ stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear;
+ stbi__gif_lzw *p;
+
+ lzw_cs = stbi__get8(s);
+ if (lzw_cs > 12) return NULL;
+ clear = 1 << lzw_cs;
+ first = 1;
+ codesize = lzw_cs + 1;
+ codemask = (1 << codesize) - 1;
+ bits = 0;
+ valid_bits = 0;
+ for (init_code = 0; init_code < clear; init_code++) {
+ g->codes[init_code].prefix = -1;
+ g->codes[init_code].first = (stbi_uc) init_code;
+ g->codes[init_code].suffix = (stbi_uc) init_code;
+ }
+
+ // support no starting clear code
+ avail = clear+2;
+ oldcode = -1;
+
+ len = 0;
+ for(;;) {
+ if (valid_bits < codesize) {
+ if (len == 0) {
+ len = stbi__get8(s); // start new block
+ if (len == 0)
+ return g->out;
+ }
+ --len;
+ bits |= (stbi__int32) stbi__get8(s) << valid_bits;
+ valid_bits += 8;
+ } else {
+ stbi__int32 code = bits & codemask;
+ bits >>= codesize;
+ valid_bits -= codesize;
+ // @OPTIMIZE: is there some way we can accelerate the non-clear path?
+ if (code == clear) { // clear code
+ codesize = lzw_cs + 1;
+ codemask = (1 << codesize) - 1;
+ avail = clear + 2;
+ oldcode = -1;
+ first = 0;
+ } else if (code == clear + 1) { // end of stream code
+ stbi__skip(s, len);
+ while ((len = stbi__get8(s)) > 0)
+ stbi__skip(s,len);
+ return g->out;
+ } else if (code <= avail) {
+ if (first) {
+ return stbi__errpuc("no clear code", "Corrupt GIF");
+ }
+
+ if (oldcode >= 0) {
+ p = &g->codes[avail++];
+ if (avail > 8192) {
+ return stbi__errpuc("too many codes", "Corrupt GIF");
+ }
+
+ p->prefix = (stbi__int16) oldcode;
+ p->first = g->codes[oldcode].first;
+ p->suffix = (code == avail) ? p->first : g->codes[code].first;
+ } else if (code == avail)
+ return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+
+ stbi__out_gif_code(g, (stbi__uint16) code);
+
+ if ((avail & codemask) == 0 && avail <= 0x0FFF) {
+ codesize++;
+ codemask = (1 << codesize) - 1;
+ }
+
+ oldcode = code;
+ } else {
+ return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+ }
+ }
+ }
+}
+
+// this function is designed to support animated gifs, although stb_image doesn't support it
+// two back is the image from two frames ago, used for a very specific disposal format
+static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back)
+{
+ int dispose;
+ int first_frame;
+ int pi;
+ int pcount;
+ STBI_NOTUSED(req_comp);
+
+ // on first frame, any non-written pixels get the background colour (non-transparent)
+ first_frame = 0;
+ if (g->out == 0) {
+ if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header
+ if (!stbi__mad3sizes_valid(4, g->w, g->h, 0))
+ return stbi__errpuc("too large", "GIF image is too large");
+ pcount = g->w * g->h;
+ g->out = (stbi_uc *) stbi__malloc(4 * pcount);
+ g->background = (stbi_uc *) stbi__malloc(4 * pcount);
+ g->history = (stbi_uc *) stbi__malloc(pcount);
+ if (!g->out || !g->background || !g->history)
+ return stbi__errpuc("outofmem", "Out of memory");
+
+ // image is treated as "transparent" at the start - ie, nothing overwrites the current background;
+ // background colour is only used for pixels that are not rendered first frame, after that "background"
+ // color refers to the color that was there the previous frame.
+ memset(g->out, 0x00, 4 * pcount);
+ memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent)
+ memset(g->history, 0x00, pcount); // pixels that were affected previous frame
+ first_frame = 1;
+ } else {
+ // second frame - how do we dispoase of the previous one?
+ dispose = (g->eflags & 0x1C) >> 2;
+ pcount = g->w * g->h;
+
+ if ((dispose == 3) && (two_back == 0)) {
+ dispose = 2; // if I don't have an image to revert back to, default to the old background
+ }
+
+ if (dispose == 3) { // use previous graphic
+ for (pi = 0; pi < pcount; ++pi) {
+ if (g->history[pi]) {
+ memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 );
+ }
+ }
+ } else if (dispose == 2) {
+ // restore what was changed last frame to background before that frame;
+ for (pi = 0; pi < pcount; ++pi) {
+ if (g->history[pi]) {
+ memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 );
+ }
+ }
+ } else {
+ // This is a non-disposal case eithe way, so just
+ // leave the pixels as is, and they will become the new background
+ // 1: do not dispose
+ // 0: not specified.
+ }
+
+ // background is what out is after the undoing of the previou frame;
+ memcpy( g->background, g->out, 4 * g->w * g->h );
+ }
+
+ // clear my history;
+ memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame
+
+ for (;;) {
+ int tag = stbi__get8(s);
+ switch (tag) {
+ case 0x2C: /* Image Descriptor */
+ {
+ stbi__int32 x, y, w, h;
+ stbi_uc *o;
+
+ x = stbi__get16le(s);
+ y = stbi__get16le(s);
+ w = stbi__get16le(s);
+ h = stbi__get16le(s);
+ if (((x + w) > (g->w)) || ((y + h) > (g->h)))
+ return stbi__errpuc("bad Image Descriptor", "Corrupt GIF");
+
+ g->line_size = g->w * 4;
+ g->start_x = x * 4;
+ g->start_y = y * g->line_size;
+ g->max_x = g->start_x + w * 4;
+ g->max_y = g->start_y + h * g->line_size;
+ g->cur_x = g->start_x;
+ g->cur_y = g->start_y;
+
+ // if the width of the specified rectangle is 0, that means
+ // we may not see *any* pixels or the image is malformed;
+ // to make sure this is caught, move the current y down to
+ // max_y (which is what out_gif_code checks).
+ if (w == 0)
+ g->cur_y = g->max_y;
+
+ g->lflags = stbi__get8(s);
+
+ if (g->lflags & 0x40) {
+ g->step = 8 * g->line_size; // first interlaced spacing
+ g->parse = 3;
+ } else {
+ g->step = g->line_size;
+ g->parse = 0;
+ }
+
+ if (g->lflags & 0x80) {
+ stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1);
+ g->color_table = (stbi_uc *) g->lpal;
+ } else if (g->flags & 0x80) {
+ g->color_table = (stbi_uc *) g->pal;
+ } else
+ return stbi__errpuc("missing color table", "Corrupt GIF");
+
+ o = stbi__process_gif_raster(s, g);
+ if (!o) return NULL;
+
+ // if this was the first frame,
+ pcount = g->w * g->h;
+ if (first_frame && (g->bgindex > 0)) {
+ // if first frame, any pixel not drawn to gets the background color
+ for (pi = 0; pi < pcount; ++pi) {
+ if (g->history[pi] == 0) {
+ g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be;
+ memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 );
+ }
+ }
+ }
+
+ return o;
+ }
+
+ case 0x21: // Comment Extension.
+ {
+ int len;
+ int ext = stbi__get8(s);
+ if (ext == 0xF9) { // Graphic Control Extension.
+ len = stbi__get8(s);
+ if (len == 4) {
+ g->eflags = stbi__get8(s);
+ g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths.
+
+ // unset old transparent
+ if (g->transparent >= 0) {
+ g->pal[g->transparent][3] = 255;
+ }
+ if (g->eflags & 0x01) {
+ g->transparent = stbi__get8(s);
+ if (g->transparent >= 0) {
+ g->pal[g->transparent][3] = 0;
+ }
+ } else {
+ // don't need transparent
+ stbi__skip(s, 1);
+ g->transparent = -1;
+ }
+ } else {
+ stbi__skip(s, len);
+ break;
+ }
+ }
+ while ((len = stbi__get8(s)) != 0) {
+ stbi__skip(s, len);
+ }
+ break;
+ }
+
+ case 0x3B: // gif stream termination code
+ return (stbi_uc *) s; // using '1' causes warning on some compilers
+
+ default:
+ return stbi__errpuc("unknown code", "Corrupt GIF");
+ }
+ }
+}
+
+static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+ if (stbi__gif_test(s)) {
+ int layers = 0;
+ stbi_uc *u = 0;
+ stbi_uc *out = 0;
+ stbi_uc *two_back = 0;
+ stbi__gif g;
+ int stride;
+ memset(&g, 0, sizeof(g));
+ if (delays) {
+ *delays = 0;
+ }
+
+ do {
+ u = stbi__gif_load_next(s, &g, comp, req_comp, two_back);
+ if (u == (stbi_uc *) s) u = 0; // end of animated gif marker
+
+ if (u) {
+ *x = g.w;
+ *y = g.h;
+ ++layers;
+ stride = g.w * g.h * 4;
+
+ if (out) {
+ out = (stbi_uc*) STBI_REALLOC( out, layers * stride );
+ if (delays) {
+ *delays = (int*) STBI_REALLOC( *delays, sizeof(int) * layers );
+ }
+ } else {
+ out = (stbi_uc*)stbi__malloc( layers * stride );
+ if (delays) {
+ *delays = (int*) stbi__malloc( layers * sizeof(int) );
+ }
+ }
+ memcpy( out + ((layers - 1) * stride), u, stride );
+ if (layers >= 2) {
+ two_back = out - 2 * stride;
+ }
+
+ if (delays) {
+ (*delays)[layers - 1U] = g.delay;
+ }
+ }
+ } while (u != 0);
+
+ // free temp buffer;
+ STBI_FREE(g.out);
+ STBI_FREE(g.history);
+ STBI_FREE(g.background);
+
+ // do the final conversion after loading everything;
+ if (req_comp && req_comp != 4)
+ out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h);
+
+ *z = layers;
+ return out;
+ } else {
+ return stbi__errpuc("not GIF", "Image was not as a gif type.");
+ }
+}
+
+static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ stbi_uc *u = 0;
+ stbi__gif g;
+ memset(&g, 0, sizeof(g));
+ STBI_NOTUSED(ri);
+
+ u = stbi__gif_load_next(s, &g, comp, req_comp, 0);
+ if (u == (stbi_uc *) s) u = 0; // end of animated gif marker
+ if (u) {
+ *x = g.w;
+ *y = g.h;
+
+ // moved conversion to after successful load so that the same
+ // can be done for multiple frames.
+ if (req_comp && req_comp != 4)
+ u = stbi__convert_format(u, 4, req_comp, g.w, g.h);
+ } else if (g.out) {
+ // if there was an error and we allocated an image buffer, free it!
+ STBI_FREE(g.out);
+ }
+
+ // free buffers needed for multiple frame loading;
+ STBI_FREE(g.history);
+ STBI_FREE(g.background);
+
+ return u;
+}
+
+static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ return stbi__gif_info_raw(s,x,y,comp);
+}
+#endif
+
+// *************************************************************************************************
+// Radiance RGBE HDR loader
+// originally by Nicolas Schulz
+#ifndef STBI_NO_HDR
+static int stbi__hdr_test_core(stbi__context *s, const char *signature)
+{
+ int i;
+ for (i=0; signature[i]; ++i)
+ if (stbi__get8(s) != signature[i])
+ return 0;
+ stbi__rewind(s);
+ return 1;
+}
+
+static int stbi__hdr_test(stbi__context* s)
+{
+ int r = stbi__hdr_test_core(s, "#?RADIANCE\n");
+ stbi__rewind(s);
+ if(!r) {
+ r = stbi__hdr_test_core(s, "#?RGBE\n");
+ stbi__rewind(s);
+ }
+ return r;
+}
+
+#define STBI__HDR_BUFLEN 1024
+static char *stbi__hdr_gettoken(stbi__context *z, char *buffer)
+{
+ int len=0;
+ char c = '\0';
+
+ c = (char) stbi__get8(z);
+
+ while (!stbi__at_eof(z) && c != '\n') {
+ buffer[len++] = c;
+ if (len == STBI__HDR_BUFLEN-1) {
+ // flush to end of line
+ while (!stbi__at_eof(z) && stbi__get8(z) != '\n')
+ ;
+ break;
+ }
+ c = (char) stbi__get8(z);
+ }
+
+ buffer[len] = 0;
+ return buffer;
+}
+
+static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp)
+{
+ if ( input[3] != 0 ) {
+ float f1;
+ // Exponent
+ f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8));
+ if (req_comp <= 2)
+ output[0] = (input[0] + input[1] + input[2]) * f1 / 3;
+ else {
+ output[0] = input[0] * f1;
+ output[1] = input[1] * f1;
+ output[2] = input[2] * f1;
+ }
+ if (req_comp == 2) output[1] = 1;
+ if (req_comp == 4) output[3] = 1;
+ } else {
+ switch (req_comp) {
+ case 4: output[3] = 1; /* fallthrough */
+ case 3: output[0] = output[1] = output[2] = 0;
+ break;
+ case 2: output[1] = 1; /* fallthrough */
+ case 1: output[0] = 0;
+ break;
+ }
+ }
+}
+
+static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ char buffer[STBI__HDR_BUFLEN];
+ char *token;
+ int valid = 0;
+ int width, height;
+ stbi_uc *scanline;
+ float *hdr_data;
+ int len;
+ unsigned char count, value;
+ int i, j, k, c1,c2, z;
+ const char *headerToken;
+ STBI_NOTUSED(ri);
+
+ // Check identifier
+ headerToken = stbi__hdr_gettoken(s,buffer);
+ if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0)
+ return stbi__errpf("not HDR", "Corrupt HDR image");
+
+ // Parse header
+ for(;;) {
+ token = stbi__hdr_gettoken(s,buffer);
+ if (token[0] == 0) break;
+ if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+ }
+
+ if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format");
+
+ // Parse width and height
+ // can't use sscanf() if we're not using stdio!
+ token = stbi__hdr_gettoken(s,buffer);
+ if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+ token += 3;
+ height = (int) strtol(token, &token, 10);
+ while (*token == ' ') ++token;
+ if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+ token += 3;
+ width = (int) strtol(token, NULL, 10);
+
+ *x = width;
+ *y = height;
+
+ if (comp) *comp = 3;
+ if (req_comp == 0) req_comp = 3;
+
+ if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0))
+ return stbi__errpf("too large", "HDR image is too large");
+
+ // Read data
+ hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0);
+ if (!hdr_data)
+ return stbi__errpf("outofmem", "Out of memory");
+
+ // Load image data
+ // image data is stored as some number of sca
+ if ( width < 8 || width >= 32768) {
+ // Read flat data
+ for (j=0; j < height; ++j) {
+ for (i=0; i < width; ++i) {
+ stbi_uc rgbe[4];
+ main_decode_loop:
+ stbi__getn(s, rgbe, 4);
+ stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp);
+ }
+ }
+ } else {
+ // Read RLE-encoded data
+ scanline = NULL;
+
+ for (j = 0; j < height; ++j) {
+ c1 = stbi__get8(s);
+ c2 = stbi__get8(s);
+ len = stbi__get8(s);
+ if (c1 != 2 || c2 != 2 || (len & 0x80)) {
+ // not run-length encoded, so we have to actually use THIS data as a decoded
+ // pixel (note this can't be a valid pixel--one of RGB must be >= 128)
+ stbi_uc rgbe[4];
+ rgbe[0] = (stbi_uc) c1;
+ rgbe[1] = (stbi_uc) c2;
+ rgbe[2] = (stbi_uc) len;
+ rgbe[3] = (stbi_uc) stbi__get8(s);
+ stbi__hdr_convert(hdr_data, rgbe, req_comp);
+ i = 1;
+ j = 0;
+ STBI_FREE(scanline);
+ goto main_decode_loop; // yes, this makes no sense
+ }
+ len <<= 8;
+ len |= stbi__get8(s);
+ if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); }
+ if (scanline == NULL) {
+ scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0);
+ if (!scanline) {
+ STBI_FREE(hdr_data);
+ return stbi__errpf("outofmem", "Out of memory");
+ }
+ }
+
+ for (k = 0; k < 4; ++k) {
+ int nleft;
+ i = 0;
+ while ((nleft = width - i) > 0) {
+ count = stbi__get8(s);
+ if (count > 128) {
+ // Run
+ value = stbi__get8(s);
+ count -= 128;
+ if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+ for (z = 0; z < count; ++z)
+ scanline[i++ * 4 + k] = value;
+ } else {
+ // Dump
+ if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+ for (z = 0; z < count; ++z)
+ scanline[i++ * 4 + k] = stbi__get8(s);
+ }
+ }
+ }
+ for (i=0; i < width; ++i)
+ stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp);
+ }
+ if (scanline)
+ STBI_FREE(scanline);
+ }
+
+ return hdr_data;
+}
+
+static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ char buffer[STBI__HDR_BUFLEN];
+ char *token;
+ int valid = 0;
+ int dummy;
+
+ if (!x) x = &dummy;
+ if (!y) y = &dummy;
+ if (!comp) comp = &dummy;
+
+ if (stbi__hdr_test(s) == 0) {
+ stbi__rewind( s );
+ return 0;
+ }
+
+ for(;;) {
+ token = stbi__hdr_gettoken(s,buffer);
+ if (token[0] == 0) break;
+ if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+ }
+
+ if (!valid) {
+ stbi__rewind( s );
+ return 0;
+ }
+ token = stbi__hdr_gettoken(s,buffer);
+ if (strncmp(token, "-Y ", 3)) {
+ stbi__rewind( s );
+ return 0;
+ }
+ token += 3;
+ *y = (int) strtol(token, &token, 10);
+ while (*token == ' ') ++token;
+ if (strncmp(token, "+X ", 3)) {
+ stbi__rewind( s );
+ return 0;
+ }
+ token += 3;
+ *x = (int) strtol(token, NULL, 10);
+ *comp = 3;
+ return 1;
+}
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ void *p;
+ stbi__bmp_data info;
+
+ info.all_a = 255;
+ p = stbi__bmp_parse_header(s, &info);
+ stbi__rewind( s );
+ if (p == NULL)
+ return 0;
+ if (x) *x = s->img_x;
+ if (y) *y = s->img_y;
+ if (comp) {
+ if (info.bpp == 24 && info.ma == 0xff000000)
+ *comp = 3;
+ else
+ *comp = info.ma ? 4 : 3;
+ }
+ return 1;
+}
+#endif
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ int channelCount, dummy, depth;
+ if (!x) x = &dummy;
+ if (!y) y = &dummy;
+ if (!comp) comp = &dummy;
+ if (stbi__get32be(s) != 0x38425053) {
+ stbi__rewind( s );
+ return 0;
+ }
+ if (stbi__get16be(s) != 1) {
+ stbi__rewind( s );
+ return 0;
+ }
+ stbi__skip(s, 6);
+ channelCount = stbi__get16be(s);
+ if (channelCount < 0 || channelCount > 16) {
+ stbi__rewind( s );
+ return 0;
+ }
+ *y = stbi__get32be(s);
+ *x = stbi__get32be(s);
+ depth = stbi__get16be(s);
+ if (depth != 8 && depth != 16) {
+ stbi__rewind( s );
+ return 0;
+ }
+ if (stbi__get16be(s) != 3) {
+ stbi__rewind( s );
+ return 0;
+ }
+ *comp = 4;
+ return 1;
+}
+
+static int stbi__psd_is16(stbi__context *s)
+{
+ int channelCount, depth;
+ if (stbi__get32be(s) != 0x38425053) {
+ stbi__rewind( s );
+ return 0;
+ }
+ if (stbi__get16be(s) != 1) {
+ stbi__rewind( s );
+ return 0;
+ }
+ stbi__skip(s, 6);
+ channelCount = stbi__get16be(s);
+ if (channelCount < 0 || channelCount > 16) {
+ stbi__rewind( s );
+ return 0;
+ }
+ (void) stbi__get32be(s);
+ (void) stbi__get32be(s);
+ depth = stbi__get16be(s);
+ if (depth != 16) {
+ stbi__rewind( s );
+ return 0;
+ }
+ return 1;
+}
+#endif
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ int act_comp=0,num_packets=0,chained,dummy;
+ stbi__pic_packet packets[10];
+
+ if (!x) x = &dummy;
+ if (!y) y = &dummy;
+ if (!comp) comp = &dummy;
+
+ if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) {
+ stbi__rewind(s);
+ return 0;
+ }
+
+ stbi__skip(s, 88);
+
+ *x = stbi__get16be(s);
+ *y = stbi__get16be(s);
+ if (stbi__at_eof(s)) {
+ stbi__rewind( s);
+ return 0;
+ }
+ if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) {
+ stbi__rewind( s );
+ return 0;
+ }
+
+ stbi__skip(s, 8);
+
+ do {
+ stbi__pic_packet *packet;
+
+ if (num_packets==sizeof(packets)/sizeof(packets[0]))
+ return 0;
+
+ packet = &packets[num_packets++];
+ chained = stbi__get8(s);
+ packet->size = stbi__get8(s);
+ packet->type = stbi__get8(s);
+ packet->channel = stbi__get8(s);
+ act_comp |= packet->channel;
+
+ if (stbi__at_eof(s)) {
+ stbi__rewind( s );
+ return 0;
+ }
+ if (packet->size != 8) {
+ stbi__rewind( s );
+ return 0;
+ }
+ } while (chained);
+
+ *comp = (act_comp & 0x10 ? 4 : 3);
+
+ return 1;
+}
+#endif
+
+// *************************************************************************************************
+// Portable Gray Map and Portable Pixel Map loader
+// by Ken Miller
+//
+// PGM: http://netpbm.sourceforge.net/doc/pgm.html
+// PPM: http://netpbm.sourceforge.net/doc/ppm.html
+//
+// Known limitations:
+// Does not support comments in the header section
+// Does not support ASCII image data (formats P2 and P3)
+// Does not support 16-bit-per-channel
+
+#ifndef STBI_NO_PNM
+
+static int stbi__pnm_test(stbi__context *s)
+{
+ char p, t;
+ p = (char) stbi__get8(s);
+ t = (char) stbi__get8(s);
+ if (p != 'P' || (t != '5' && t != '6')) {
+ stbi__rewind( s );
+ return 0;
+ }
+ return 1;
+}
+
+static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+ stbi_uc *out;
+ STBI_NOTUSED(ri);
+
+ if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n))
+ return 0;
+
+ *x = s->img_x;
+ *y = s->img_y;
+ if (comp) *comp = s->img_n;
+
+ if (!stbi__mad3sizes_valid(s->img_n, s->img_x, s->img_y, 0))
+ return stbi__errpuc("too large", "PNM too large");
+
+ out = (stbi_uc *) stbi__malloc_mad3(s->img_n, s->img_x, s->img_y, 0);
+ if (!out) return stbi__errpuc("outofmem", "Out of memory");
+ stbi__getn(s, out, s->img_n * s->img_x * s->img_y);
+
+ if (req_comp && req_comp != s->img_n) {
+ out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y);
+ if (out == NULL) return out; // stbi__convert_format frees input on failure
+ }
+ return out;
+}
+
+static int stbi__pnm_isspace(char c)
+{
+ return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r';
+}
+
+static void stbi__pnm_skip_whitespace(stbi__context *s, char *c)
+{
+ for (;;) {
+ while (!stbi__at_eof(s) && stbi__pnm_isspace(*c))
+ *c = (char) stbi__get8(s);
+
+ if (stbi__at_eof(s) || *c != '#')
+ break;
+
+ while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' )
+ *c = (char) stbi__get8(s);
+ }
+}
+
+static int stbi__pnm_isdigit(char c)
+{
+ return c >= '0' && c <= '9';
+}
+
+static int stbi__pnm_getinteger(stbi__context *s, char *c)
+{
+ int value = 0;
+
+ while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) {
+ value = value*10 + (*c - '0');
+ *c = (char) stbi__get8(s);
+ }
+
+ return value;
+}
+
+static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp)
+{
+ int maxv, dummy;
+ char c, p, t;
+
+ if (!x) x = &dummy;
+ if (!y) y = &dummy;
+ if (!comp) comp = &dummy;
+
+ stbi__rewind(s);
+
+ // Get identifier
+ p = (char) stbi__get8(s);
+ t = (char) stbi__get8(s);
+ if (p != 'P' || (t != '5' && t != '6')) {
+ stbi__rewind(s);
+ return 0;
+ }
+
+ *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm
+
+ c = (char) stbi__get8(s);
+ stbi__pnm_skip_whitespace(s, &c);
+
+ *x = stbi__pnm_getinteger(s, &c); // read width
+ stbi__pnm_skip_whitespace(s, &c);
+
+ *y = stbi__pnm_getinteger(s, &c); // read height
+ stbi__pnm_skip_whitespace(s, &c);
+
+ maxv = stbi__pnm_getinteger(s, &c); // read max value
+
+ if (maxv > 255)
+ return stbi__err("max value > 255", "PPM image not 8-bit");
+ else
+ return 1;
+}
+#endif
+
+static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)
+{
+ #ifndef STBI_NO_JPEG
+ if (stbi__jpeg_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_PNG
+ if (stbi__png_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_GIF
+ if (stbi__gif_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_BMP
+ if (stbi__bmp_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_PSD
+ if (stbi__psd_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_PIC
+ if (stbi__pic_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_PNM
+ if (stbi__pnm_info(s, x, y, comp)) return 1;
+ #endif
+
+ #ifndef STBI_NO_HDR
+ if (stbi__hdr_info(s, x, y, comp)) return 1;
+ #endif
+
+ // test tga last because it's a crappy test!
+ #ifndef STBI_NO_TGA
+ if (stbi__tga_info(s, x, y, comp))
+ return 1;
+ #endif
+ return stbi__err("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static int stbi__is_16_main(stbi__context *s)
+{
+ #ifndef STBI_NO_PNG
+ if (stbi__png_is16(s)) return 1;
+ #endif
+
+ #ifndef STBI_NO_PSD
+ if (stbi__psd_is16(s)) return 1;
+ #endif
+
+ return 0;
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp)
+{
+ FILE *f = stbi__fopen(filename, "rb");
+ int result;
+ if (!f) return stbi__err("can't fopen", "Unable to open file");
+ result = stbi_info_from_file(f, x, y, comp);
+ fclose(f);
+ return result;
+}
+
+STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp)
+{
+ int r;
+ stbi__context s;
+ long pos = ftell(f);
+ stbi__start_file(&s, f);
+ r = stbi__info_main(&s,x,y,comp);
+ fseek(f,pos,SEEK_SET);
+ return r;
+}
+
+STBIDEF int stbi_is_16_bit(char const *filename)
+{
+ FILE *f = stbi__fopen(filename, "rb");
+ int result;
+ if (!f) return stbi__err("can't fopen", "Unable to open file");
+ result = stbi_is_16_bit_from_file(f);
+ fclose(f);
+ return result;
+}
+
+STBIDEF int stbi_is_16_bit_from_file(FILE *f)
+{
+ int r;
+ stbi__context s;
+ long pos = ftell(f);
+ stbi__start_file(&s, f);
+ r = stbi__is_16_main(&s);
+ fseek(f,pos,SEEK_SET);
+ return r;
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp)
+{
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp)
+{
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+ return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len)
+{
+ stbi__context s;
+ stbi__start_mem(&s,buffer,len);
+ return stbi__is_16_main(&s);
+}
+
+STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user)
+{
+ stbi__context s;
+ stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+ return stbi__is_16_main(&s);
+}
+
+#endif // STB_IMAGE_IMPLEMENTATION
+
+/*
+ revision history:
+ 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs
+ 2.19 (2018-02-11) fix warning
+ 2.18 (2018-01-30) fix warnings
+ 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug
+ 1-bit BMP
+ *_is_16_bit api
+ avoid warnings
+ 2.16 (2017-07-23) all functions have 16-bit variants;
+ STBI_NO_STDIO works again;
+ compilation fixes;
+ fix rounding in unpremultiply;
+ optimize vertical flip;
+ disable raw_len validation;
+ documentation fixes
+ 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode;
+ warning fixes; disable run-time SSE detection on gcc;
+ uniform handling of optional "return" values;
+ thread-safe initialization of zlib tables
+ 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+ 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now
+ 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+ 2.11 (2016-04-02) allocate large structures on the stack
+ remove white matting for transparent PSD
+ fix reported channel count for PNG & BMP
+ re-enable SSE2 in non-gcc 64-bit
+ support RGB-formatted JPEG
+ read 16-bit PNGs (only as 8-bit)
+ 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED
+ 2.09 (2016-01-16) allow comments in PNM files
+ 16-bit-per-pixel TGA (not bit-per-component)
+ info() for TGA could break due to .hdr handling
+ info() for BMP to shares code instead of sloppy parse
+ can use STBI_REALLOC_SIZED if allocator doesn't support realloc
+ code cleanup
+ 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA
+ 2.07 (2015-09-13) fix compiler warnings
+ partial animated GIF support
+ limited 16-bpc PSD support
+ #ifdef unused functions
+ bug with < 92 byte PIC,PNM,HDR,TGA
+ 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value
+ 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning
+ 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit
+ 2.03 (2015-04-12) extra corruption checking (mmozeiko)
+ stbi_set_flip_vertically_on_load (nguillemot)
+ fix NEON support; fix mingw support
+ 2.02 (2015-01-19) fix incorrect assert, fix warning
+ 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2
+ 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG
+ 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg)
+ progressive JPEG (stb)
+ PGM/PPM support (Ken Miller)
+ STBI_MALLOC,STBI_REALLOC,STBI_FREE
+ GIF bugfix -- seemingly never worked
+ STBI_NO_*, STBI_ONLY_*
+ 1.48 (2014-12-14) fix incorrectly-named assert()
+ 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb)
+ optimize PNG (ryg)
+ fix bug in interlaced PNG with user-specified channel count (stb)
+ 1.46 (2014-08-26)
+ fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG
+ 1.45 (2014-08-16)
+ fix MSVC-ARM internal compiler error by wrapping malloc
+ 1.44 (2014-08-07)
+ various warning fixes from Ronny Chevalier
+ 1.43 (2014-07-15)
+ fix MSVC-only compiler problem in code changed in 1.42
+ 1.42 (2014-07-09)
+ don't define _CRT_SECURE_NO_WARNINGS (affects user code)
+ fixes to stbi__cleanup_jpeg path
+ added STBI_ASSERT to avoid requiring assert.h
+ 1.41 (2014-06-25)
+ fix search&replace from 1.36 that messed up comments/error messages
+ 1.40 (2014-06-22)
+ fix gcc struct-initialization warning
+ 1.39 (2014-06-15)
+ fix to TGA optimization when req_comp != number of components in TGA;
+ fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite)
+ add support for BMP version 5 (more ignored fields)
+ 1.38 (2014-06-06)
+ suppress MSVC warnings on integer casts truncating values
+ fix accidental rename of 'skip' field of I/O
+ 1.37 (2014-06-04)
+ remove duplicate typedef
+ 1.36 (2014-06-03)
+ convert to header file single-file library
+ if de-iphone isn't set, load iphone images color-swapped instead of returning NULL
+ 1.35 (2014-05-27)
+ various warnings
+ fix broken STBI_SIMD path
+ fix bug where stbi_load_from_file no longer left file pointer in correct place
+ fix broken non-easy path for 32-bit BMP (possibly never used)
+ TGA optimization by Arseny Kapoulkine
+ 1.34 (unknown)
+ use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case
+ 1.33 (2011-07-14)
+ make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements
+ 1.32 (2011-07-13)
+ support for "info" function for all supported filetypes (SpartanJ)
+ 1.31 (2011-06-20)
+ a few more leak fixes, bug in PNG handling (SpartanJ)
+ 1.30 (2011-06-11)
+ added ability to load files via callbacks to accomidate custom input streams (Ben Wenger)
+ removed deprecated format-specific test/load functions
+ removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway
+ error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha)
+ fix inefficiency in decoding 32-bit BMP (David Woo)
+ 1.29 (2010-08-16)
+ various warning fixes from Aurelien Pocheville
+ 1.28 (2010-08-01)
+ fix bug in GIF palette transparency (SpartanJ)
+ 1.27 (2010-08-01)
+ cast-to-stbi_uc to fix warnings
+ 1.26 (2010-07-24)
+ fix bug in file buffering for PNG reported by SpartanJ
+ 1.25 (2010-07-17)
+ refix trans_data warning (Won Chun)
+ 1.24 (2010-07-12)
+ perf improvements reading from files on platforms with lock-heavy fgetc()
+ minor perf improvements for jpeg
+ deprecated type-specific functions so we'll get feedback if they're needed
+ attempt to fix trans_data warning (Won Chun)
+ 1.23 fixed bug in iPhone support
+ 1.22 (2010-07-10)
+ removed image *writing* support
+ stbi_info support from Jetro Lauha
+ GIF support from Jean-Marc Lienher
+ iPhone PNG-extensions from James Brown
+ warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva)
+ 1.21 fix use of 'stbi_uc' in header (reported by jon blow)
+ 1.20 added support for Softimage PIC, by Tom Seddon
+ 1.19 bug in interlaced PNG corruption check (found by ryg)
+ 1.18 (2008-08-02)
+ fix a threading bug (local mutable static)
+ 1.17 support interlaced PNG
+ 1.16 major bugfix - stbi__convert_format converted one too many pixels
+ 1.15 initialize some fields for thread safety
+ 1.14 fix threadsafe conversion bug
+ header-file-only version (#define STBI_HEADER_FILE_ONLY before including)
+ 1.13 threadsafe
+ 1.12 const qualifiers in the API
+ 1.11 Support installable IDCT, colorspace conversion routines
+ 1.10 Fixes for 64-bit (don't use "unsigned long")
+ optimized upsampling by Fabian "ryg" Giesen
+ 1.09 Fix format-conversion for PSD code (bad global variables!)
+ 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz
+ 1.07 attempt to fix C++ warning/errors again
+ 1.06 attempt to fix C++ warning/errors again
+ 1.05 fix TGA loading to return correct *comp and use good luminance calc
+ 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free
+ 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR
+ 1.02 support for (subset of) HDR files, float interface for preferred access to them
+ 1.01 fix bug: possible bug in handling right-side up bmps... not sure
+ fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all
+ 1.00 interface to zlib that skips zlib header
+ 0.99 correct handling of alpha in palette
+ 0.98 TGA loader by lonesock; dynamically add loaders (untested)
+ 0.97 jpeg errors on too large a file; also catch another malloc failure
+ 0.96 fix detection of invalid v value - particleman@mollyrocket forum
+ 0.95 during header scan, seek to markers in case of padding
+ 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same
+ 0.93 handle jpegtran output; verbose errors
+ 0.92 read 4,8,16,24,32-bit BMP files of several formats
+ 0.91 output 24-bit Windows 3.0 BMP files
+ 0.90 fix a few more warnings; bump version number to approach 1.0
+ 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd
+ 0.60 fix compiling as c++
+ 0.59 fix warnings: merge Dave Moore's -Wall fixes
+ 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian
+ 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available
+ 0.56 fix bug: zlib uncompressed mode len vs. nlen
+ 0.55 fix bug: restart_interval not initialized to 0
+ 0.54 allow NULL for 'int *comp'
+ 0.53 fix bug in png 3->4; speedup png decoding
+ 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments
+ 0.51 obey req_comp requests, 1-component jpegs return as 1-component,
+ on 'test' only check type, not whether we support this variant
+ 0.50 (2006-11-19)
+ first released version
+*/
+
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this
+software, either in source code form or as a compiled binary, for any purpose,
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this
+software dedicate any and all copyright interest in the software to the public
+domain. We make this dedication for the benefit of the public at large and to
+the detriment of our heirs and successors. We intend this dedication to be an
+overt act of relinquishment in perpetuity of all present and future rights to
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/